arm64 kexec kernel patches v12
Subsystems (7)
v12 cover arm64 kexec kernel patches v12
Hi All,
This series adds the core support for kexec re-boot and kdump on ARM64. This version of the series combines Takahiro's kdump patches with my kexec patches.
Changes since v11:
o No changes, rebase to v4.4-rc2.
Changes since v10:
o Move the flush of the new image from arm64_relocate_new_kernel to machine_kexec.
o Pass values to arm64_relocate_new_kernel in registers, not in global variables.
o Fixups to setting the sctlr_el1 and sctlr_el2 flags.
To load a second stage kernel and execute a kexec re-boot or to work with kdump on ARM64 systems a series of patches to kexec-tools [2], which have not yet been merged upstream, are needed.
I have tested kexec with the ARM Foundation model, and Takahiro has reported that kdump is working on the 96boards HiKey developer board. Kexec on EFI systems works correctly. More ACPI + kexec testing is needed.
Patch 1 here moves the macros from proc-macros.S to asm/assembler.h so that the dcache_line_size macro it defines can be uesd by kexec's relocate kernel routine.
Patches 2 & 3 rework the ARM64 hcall mechanism to give the CPU reset routines the ability to switch exception levels from EL1 to EL2 for kernels that were entered in EL2.
Patch 4 allows KVM to handle a CPU reset.
Patches 5 - 7 add back the ARM64 CPU reset support that was recently removed from the kernel.
Patches 8 - 10 add kexec support.
Patches 11-16 add kdump support.
Please consider all patches for inclusion.
[1] https://git.kernel.org/cgit/linux/kernel/git/geoff/linux-kexec.git
[2] https://git.kernel.org/cgit/linux/kernel/git/geoff/kexec-tools.git
-Geoff
The following changes since commit 1ec218373b8ebda821aec00bb156a9c94fad9cd4:
Linux 4.4-rc2 (2015-11-22 16:45:59 -0800)
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/geoff/linux-kexec.git kexec-v12
for you to fetch changes up to 83db66df70fc1f7cd4da8b580d264b9320991cf0:
arm64: kdump: relax BUG_ON() if more than one cpus are still active (2015-11-24 13:59:27 -0800)
----------------------------------------------------------------
AKASHI Takahiro (7):
arm64: kvm: allows kvm cpu hotplug
arm64: kdump: reserve memory for crash dump kernel
arm64: kdump: implement machine_crash_shutdown()
arm64: kdump: add kdump support
arm64: kdump: update a kernel doc
arm64: kdump: enable kdump in the arm64 defconfig
arm64: kdump: relax BUG_ON() if more than one cpus are still active
Geoff Levand (9):
arm64: Fold proc-macros.S into assembler.h
arm64: Convert hcalls to use HVC immediate value
arm64: Add new hcall HVC_CALL_FUNC
arm64: Add back cpu_reset routines
Revert "arm64: mm: remove unused cpu_set_idmap_tcr_t0sz function" Revert "arm64: remove dead code"
arm64/kexec: Add core kexec support
arm64/kexec: Add pr_devel output
arm64/kexec: Enable kexec in the arm64 defconfig
Documentation/kdump/kdump.txt | 32 ++++- arch/arm/include/asm/kvm_host.h | 10 +- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++----- arch/arm/kvm/mmu.c | 5 + arch/arm64/Kconfig | 22 ++++ arch/arm64/configs/defconfig | 2 + arch/arm64/include/asm/assembler.h | 48 ++++++- arch/arm64/include/asm/kexec.h | 80 +++++++++++ arch/arm64/include/asm/kvm_host.h | 16 ++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/mmu.h | 1 + arch/arm64/include/asm/mmu_context.h | 35 +++-- arch/arm64/include/asm/virt.h | 49 +++++++ arch/arm64/kernel/Makefile | 3 + arch/arm64/kernel/cpu-reset.S | 84 ++++++++++++ arch/arm64/kernel/cpu-reset.h | 22 ++++ arch/arm64/kernel/crash_dump.c | 71 ++++++++++ arch/arm64/kernel/hyp-stub.S | 43 ++++-- arch/arm64/kernel/machine_kexec.c | 248 +++++++++++++++++++++++++++++++++++ arch/arm64/kernel/relocate_kernel.S | 131 ++++++++++++++++++ arch/arm64/kernel/setup.c | 7 +- arch/arm64/kernel/smp.c | 16 ++- arch/arm64/kvm/hyp-init.S | 34 ++++- arch/arm64/kvm/hyp.S | 44 +++++-- arch/arm64/mm/cache.S | 2 - arch/arm64/mm/init.c | 83 ++++++++++++ arch/arm64/mm/mmu.c | 11 ++ arch/arm64/mm/proc-macros.S | 64 --------- arch/arm64/mm/proc.S | 3 - include/uapi/linux/kexec.h | 1 + 31 files changed, 1097 insertions(+), 151 deletions(-) create mode 100644 arch/arm64/include/asm/kexec.h create mode 100644 arch/arm64/kernel/cpu-reset.S create mode 100644 arch/arm64/kernel/cpu-reset.h create mode 100644 arch/arm64/kernel/crash_dump.c create mode 100644 arch/arm64/kernel/machine_kexec.c create mode 100644 arch/arm64/kernel/relocate_kernel.S delete mode 100644 arch/arm64/mm/proc-macros.S
Signature · 7 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 1/16 arm64: Fold proc-macros.S into assembler.h
To allow the assembler macros defined in arch/arm64/mm/proc-macros.S to be used outside the mm code move the contents of proc-macros.S to asm/assembler.h. Also, delete proc-macros.S, and fix up all references to proc-macros.S.
--- arch/arm64/include/asm/assembler.h | 48 +++++++++++++++++++++++++++- arch/arm64/kvm/hyp-init.S | 1 - arch/arm64/mm/cache.S | 2 -- arch/arm64/mm/proc-macros.S | 64 -------------------------------------- arch/arm64/mm/proc.S | 3 -- 5 files changed, 47 insertions(+), 71 deletions(-) delete mode 100644 arch/arm64/mm/proc-macros.S
arch/arm64/include/asm/assembler.h +47 -1
@@ -1,5 +1,5 @@/* -* Based on arch/arm/include/asm/assembler.h +* Based on arch/arm/include/asm/assembler.h, arch/arm/mm/proc-macros.S * * Copyright (C) 1996-2000 Russell King * Copyright (C) 2012 ARM Ltd.
@@ -23,6 +23,8 @@#ifndef __ASM_ASSEMBLER_H#define __ASM_ASSEMBLER_H +#include <asm/asm-offsets.h> +#include <asm/pgtable-hwdef.h> #include <asm/ptrace.h> #include <asm/thread_info.h>
@@ -194,6 +196,50 @@ lr .req x30 // link register .endm /* + * vma_vm_mm - get mm pointer from vma pointer (vma->vm_mm) + */ + .macro vma_vm_mm, rd, rn + ldr \rd, [\rn, #VMA_VM_MM] + .endm + +/* +* mmid - get context id from mm pointer (mm->context.id) +*/ + .macro mmid, rd, rn + ldr \rd, [\rn, #MM_CONTEXT_ID] + .endm + +/* +* dcache_line_size - get the minimum D-cache line size from the CTR register. +*/ + .macro dcache_line_size, reg, tmp + mrs \tmp, ctr_el0 // read CTR + ubfm \tmp, \tmp, #16, #19 // cache line size encoding + mov \reg, #4 // bytes per word + lsl \reg, \reg, \tmp // actual cache line size + .endm + +/* + * icache_line_size - get the minimum I-cache line size from the CTR register. + */ + .macro icache_line_size, reg, tmp + mrs \tmp, ctr_el0 // read CTR + and \tmp, \tmp, #0xf // cache line size encoding + mov \reg, #4 // bytes per word + lsl \reg, \reg, \tmp // actual cache line size + .endm + +/* + * tcr_set_idmap_t0sz - update TCR.T0SZ so that we can load the ID map + */ + .macro tcr_set_idmap_t0sz, valreg, tmpreg +#ifndef CONFIG_ARM64_VA_BITS_48 + ldr_l \tmpreg, idmap_t0sz + bfi \valreg, \tmpreg, #TCR_T0SZ_OFFSET, #TCR_TxSZ_WIDTH +#endif + .endm + +/* * Annotate a function as position independent, i.e., safe to be called before * the kernel virtual mapping is activated. */
arch/arm64/kvm/hyp-init.S +0 -1
@@ -20,7 +20,6 @@ #include <asm/assembler.h> #include <asm/kvm_arm.h> #include <asm/kvm_mmu.h> -#include <asm/pgtable-hwdef.h> .text .pushsection .hyp.idmap.text, "ax"
arch/arm64/mm/cache.S +0 -2
@@ -24,8 +24,6 @@ #include <asm/cpufeature.h> #include <asm/alternative.h> -#include "proc-macros.S" - /* * flush_icache_range(start,end) *
arch/arm64/mm/proc-macros.S deleted +0 -64
64 lines not on this page · raw message
arch/arm64/mm/proc.S +0 -3
@@ -23,11 +23,8 @@ #include <asm/assembler.h> #include <asm/asm-offsets.h> #include <asm/hwcap.h> -#include <asm/pgtable-hwdef.h> #include <asm/pgtable.h> -#include "proc-macros.S" - #ifdef CONFIG_ARM64_64K_PAGES #define TCR_TG_FLAGS TCR_TG0_64K | TCR_TG1_64K #elif defined(CONFIG_ARM64_16K_PAGES)
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 2/16 arm64: Convert hcalls to use HVC immediate value
In reply to Geoff LevandThe existing arm64 hcall implementations are limited in that they only allow for two distinct hcalls; with the x0 register either zero or not zero. Also, the API of the hyp-stub exception vector routines and the KVM exception vector routines differ; hyp-stub uses a non-zero value in x0 to implement __hyp_set_vectors, whereas KVM uses it to implement kvm_call_hyp.
To allow for additional hcalls to be defined and to make the arm64 hcall API more consistent across exception vector routines, change the hcall implementations to use the 16 bit immediate value of the HVC instruction to specify the hcall type.
Define three new preprocessor macros HVC_CALL_HYP, HVC_GET_VECTORS, and HVC_SET_VECTORS to be used as hcall type specifiers and convert the existing __hyp_get_vectors(), __hyp_set_vectors() and kvm_call_hyp() routines to use these new macros when executing an HVC call. Also, change the corresponding hyp-stub and KVM el1_sync exception vector routines to use these new macros.
--- arch/arm64/include/asm/virt.h | 27 +++++++++++++++++++++++++++ arch/arm64/kernel/hyp-stub.S | 32 +++++++++++++++++++++----------- arch/arm64/kvm/hyp.S | 16 +++++++++------- 3 files changed, 57 insertions(+), 18 deletions(-)
arch/arm64/include/asm/virt.h +27 -0
@@ -18,6 +18,33 @@#ifndef __ASM__VIRT_H#define __ASM__VIRT_H +/* +* The arm64 hcall implementation uses the ISS field of the ESR_EL2 register to +* specify the hcall type. The exception handlers are allowed to use registers +* x17 and x18 in their implementation. Any routine issuing an hcall must not +* expect these registers to be preserved. +*/ + +/* +* HVC_CALL_HYP - Execute a hyp routine. +*/ + +#define HVC_CALL_HYP 0 + +/* +* HVC_GET_VECTORS - Return the value of the vbar_el2 register. +*/ + +#define HVC_GET_VECTORS 1 + +/* +* HVC_SET_VECTORS - Set the value of the vbar_el2 register. +* +* @x0: Physical address of the new vector table. +*/ + +#define HVC_SET_VECTORS 2 +#define BOOT_CPU_MODE_EL1 (0xe11)#define BOOT_CPU_MODE_EL2 (0xe12)
arch/arm64/kernel/hyp-stub.S +21 -11
@@ -22,6 +22,7 @@ #include <linux/irqchip/arm-gic-v3.h> #include <asm/assembler.h> +#include <asm/kvm_arm.h> #include <asm/ptrace.h> #include <asm/virt.h>
@@ -53,14 +54,22 @@ ENDPROC(__hyp_stub_vectors) .align 11 el1_sync: - mrs x1, esr_el2 - lsr x1, x1, #26 - cmp x1, #0x16 + mrs x18, esr_el2 + lsr x17, x18, #ESR_ELx_EC_SHIFT + and x18, x18, #ESR_ELx_ISS_MASK + + cmp x17, #ESR_ELx_EC_HVC64 b.ne 2f // Not an HVC trap - cbz x0, 1f - msr vbar_el2, x0 // Set vbar_el2 + + cmp x18, #HVC_GET_VECTORS + b.ne 1f + mrs x0, vbar_el2 b 2f -1: mrs x0, vbar_el2 // Return vbar_el2 + +1: cmp x18, #HVC_SET_VECTORS + b.ne 2f + msr vbar_el2, x0 + 2: eret ENDPROC(el1_sync)
@@ -100,11 +109,12 @@ ENDPROC(\label) * initialisation entry point. */ -ENTRY(__hyp_get_vectors) - mov x0, xzr - // fall through ENTRY(__hyp_set_vectors) - hvc #0 + hvc #HVC_SET_VECTORS ret -ENDPROC(__hyp_get_vectors) ENDPROC(__hyp_set_vectors) + +ENTRY(__hyp_get_vectors) + hvc #HVC_GET_VECTORS + ret +ENDPROC(__hyp_get_vectors)
arch/arm64/kvm/hyp.S +9 -7
@@ -29,6 +29,7 @@ #include <asm/kvm_asm.h> #include <asm/kvm_mmu.h> #include <asm/memory.h> +#include <asm/virt.h> #define CPU_GP_REG_OFFSET(x) (CPU_GP_REGS + x) #define CPU_XREG_OFFSET(x) CPU_GP_REG_OFFSET(CPU_USER_PT_REGS + 8*x)
@@ -932,12 +933,9 @@ __hyp_panic_str: * in Hyp mode (see init_hyp_mode in arch/arm/kvm/arm.c). Return values are * passed in r0 and r1. * - * A function pointer with a value of 0 has a special meaning, and is - * used to implement __hyp_get_vectors in the same way as in - * arch/arm64/kernel/hyp_stub.S. */ ENTRY(kvm_call_hyp) - hvc #0 + hvc #HVC_CALL_HYP ret ENDPROC(kvm_call_hyp)
@@ -968,6 +966,7 @@ el1_sync: // Guest trapped into EL2 mrs x1, esr_el2 lsr x2, x1, #ESR_ELx_EC_SHIFT + and x0, x1, #ESR_ELx_ISS_MASK cmp x2, #ESR_ELx_EC_HVC64 b.ne el1_trap
@@ -976,15 +975,18 @@ el1_sync: // Guest trapped into EL2 cbnz x3, el1_trap // called HVC /* Here, we're pretty sure the host called HVC. */ + mov x18, x0 pop x2, x3 pop x0, x1 - /* Check for __hyp_get_vectors */ - cbnz x0, 1f + cmp x18, #HVC_GET_VECTORS + b.ne 1f mrs x0, vbar_el2 b 2f -1: push lr, xzr +1: /* Default to HVC_CALL_HYP. */ + + push lr, xzr /* * Compute the function address in EL2, and shuffle the parameters.
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 3/16 arm64: Add new hcall HVC_CALL_FUNC
In reply to Geoff LevandAdd the new hcall HVC_CALL_FUNC that allows execution of a function at EL2. During CPU reset the CPU must be brought to the exception level it had on entry to the kernel. The HVC_CALL_FUNC hcall will provide the mechanism needed for this exception level switch.
To allow the HVC_CALL_FUNC exception vector to work without a stack, which is needed to support an hcall at CPU reset, this implementation uses register x18 to store the link register across the caller provided function. This dictates that the caller provided function must preserve the contents of register x18.
--- arch/arm64/include/asm/virt.h | 13 +++++++++++++ arch/arm64/kernel/hyp-stub.S | 13 ++++++++++++- 2 files changed, 25 insertions(+), 1 deletion(-)
arch/arm64/include/asm/virt.h +13 -0
@@ -45,6 +45,19 @@#define HVC_SET_VECTORS 2 +/* +* HVC_CALL_FUNC - Execute a function at EL2. +* +* @x0: Physical address of the function to be executed. +* @x1: Passed as the first argument to the function. +* @x2: Passed as the second argument to the function. +* @x3: Passed as the third argument to the function. +* +* The called function must preserve the contents of register x18. +*/ + +#define HVC_CALL_FUNC 3 +#define BOOT_CPU_MODE_EL1 (0xe11)#define BOOT_CPU_MODE_EL2 (0xe12)
arch/arm64/kernel/hyp-stub.S +12 -1
@@ -67,8 +67,19 @@ el1_sync: b 2f 1: cmp x18, #HVC_SET_VECTORS - b.ne 2f + b.ne 1f msr vbar_el2, x0 + b 2f + +1: cmp x18, #HVC_CALL_FUNC + b.ne 2f + mov x18, lr + mov lr, x0 + mov x0, x1 + mov x1, x2 + mov x2, x3 + blr lr + mov lr, x18 2: eret ENDPROC(el1_sync)
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 4/16 arm64: kvm: allows kvm cpu hotplug
In reply to Geoff LevandFrom: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function, kvm_arch_hardware_disable() and kvm_arch_hardware_enable() respectively.
We don't need arm64-specific cpu hotplug hook any more.
Since this patch modifies common part of code between arm and arm64, one stub definition, __cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.
--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)
arch/arm/include/asm/kvm_host.h +9 -1
@@ -214,6 +214,15 @@ static inline void __cpu_init_hyp_mode(phys_addr_t boot_pgd_ptr,kvm_call_hyp ((void *)hyp_stack_ptr, vector_ptr, pgd_ptr); } +static inline void __cpu_reset_hyp_mode (phys_addr_t boot_pgd_ptr, +phys_addr_t phys_idmap_start) +{ +/* +* TODO +* kvm_call_reset(boot_pgd_ptr, phys_idmap_start); +*/ +} +static inline int kvm_arch_dev_ioctl_check_extension (long ext) {return 0 ;
@@ -226,7 +235,6 @@ void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot);struct kvm_vcpu *kvm_mpidr_to_vcpu (struct kvm *kvm,unsigned long mpidr); -static inline void kvm_arch_hardware_disable (void ) {}static inline void kvm_arch_hardware_unsetup (void ) {}static inline void kvm_arch_sync_events (struct kvm *kvm) {}static inline void kvm_arch_vcpu_uninit (struct kvm_vcpu *vcpu) {}
arch/arm/include/asm/kvm_mmu.h +1 -0
@@ -66,6 +66,7 @@ void kvm_mmu_free_memory_caches(struct kvm_vcpu *vcpu);phys_addr_t kvm_mmu_get_httbr (void );phys_addr_t kvm_mmu_get_boot_httbr (void );phys_addr_t kvm_get_idmap_vector (void ); +phys_addr_t kvm_get_idmap_start (void );int kvm_mmu_init (void );void kvm_clear_hyp_idmap (void );
arch/arm/kvm/arm.c +37 -42
149 lines not on this page · raw message
arch/arm/kvm/mmu.c +5 -0
@@ -1644,6 +1644,11 @@ phys_addr_t kvm_get_idmap_vector(void)return hyp_idmap_vector; } +phys_addr_t kvm_get_idmap_start (void ) +{ +return hyp_idmap_start; +} +int kvm_mmu_init (void ) {int err;
arch/arm64/include/asm/kvm_host.h +15 -1
@@ -223,6 +223,7 @@ struct kvm_vcpu *kvm_arm_get_running_vcpu(void);struct kvm_vcpu *__percpu *kvm_get_running_vcpus (void );u64 kvm_call_hyp (void *hypfn, ...); +void kvm_call_reset (phys_addr_t boot_pgd_ptr,phys_addr_t phys_idmap_start);void force_vm_exit (const cpumask_t *mask);void kvm_mmu_wp_memory_region (struct kvm *kvm,int slot);
@@ -247,7 +248,20 @@ static inline void __cpu_init_hyp_mode(phys_addr_t boot_pgd_ptr, hyp_stack_ptr, vector_ptr); } -static inline void kvm_arch_hardware_disable (void ) {} +static inline void __cpu_reset_hyp_mode (phys_addr_t boot_pgd_ptr, +phys_addr_t phys_idmap_start) +{ +/* +* Call reset code, and switch back to stub hyp vectors. +*/ +kvm_call_reset (boot_pgd_ptr, phys_idmap_start); +} + +struct vgic_sr_vectors { +void *save_vgic; +void *restore_vgic; +}; +static inline void kvm_arch_hardware_unsetup (void ) {}static inline void kvm_arch_sync_events (struct kvm *kvm) {}static inline void kvm_arch_vcpu_uninit (struct kvm_vcpu *vcpu) {}
arch/arm64/include/asm/kvm_mmu.h +1 -0
@@ -98,6 +98,7 @@ void kvm_mmu_free_memory_caches(struct kvm_vcpu *vcpu);phys_addr_t kvm_mmu_get_httbr (void );phys_addr_t kvm_mmu_get_boot_httbr (void );phys_addr_t kvm_get_idmap_vector (void ); +phys_addr_t kvm_get_idmap_start (void );int kvm_mmu_init (void );void kvm_clear_hyp_idmap (void );
arch/arm64/include/asm/virt.h +9 -0
@@ -58,9 +58,18 @@#define HVC_CALL_FUNC 3 +/* +* HVC_RESET_CPU - Reset cpu in EL2 to initial state. +* +* @x0: entry address in trampoline code in va +* @x1: identical mapping page table in pa +*/ +#define BOOT_CPU_MODE_EL1 (0xe11)#define BOOT_CPU_MODE_EL2 (0xe12) +#define HVC_RESET_CPU 4 +#ifndef __ASSEMBLY__/*
arch/arm64/kvm/hyp-init.S +33 -0
@@ -139,6 +139,39 @@ merged: eret ENDPROC(__kvm_hyp_init) + /* + * x0: HYP boot pgd + * x1: HYP phys_idmap_start + */ +ENTRY(__kvm_hyp_reset) + /* We're in trampoline code in VA, switch back to boot page tables */ + msr ttbr0_el2, x0 + isb + + /* Invalidate the old TLBs */ + tlbi alle2 + dsb sy + + /* Branch into PA space */ + adr x0, 1f + bfi x1, x0, #0, #PAGE_SHIFT + br x1 + + /* We're now in idmap, disable MMU */ +1: mrs x0, sctlr_el2 + ldr x1, =SCTLR_EL2_FLAGS + bic x0, x0, x1 // Clear SCTL_M and etc + msr sctlr_el2, x0 + isb + + /* Install stub vectors */ + adrp x0, __hyp_stub_vectors + add x0, x0, #:lo12:__hyp_stub_vectors + msr vbar_el2, x0 + + eret +ENDPROC(__kvm_hyp_reset) + .ltorg .popsection
arch/arm64/kvm/hyp.S +28 -4
@@ -939,6 +939,11 @@ ENTRY(kvm_call_hyp) ret ENDPROC(kvm_call_hyp) +ENTRY(kvm_call_reset) + hvc #HVC_RESET_CPU + ret +ENDPROC(kvm_call_reset) + .macro invalid_vector label, target .align 2 \label:
@@ -982,10 +987,27 @@ el1_sync: // Guest trapped into EL2 cmp x18, #HVC_GET_VECTORS b.ne 1f mrs x0, vbar_el2 - b 2f - -1: /* Default to HVC_CALL_HYP. */ + b do_eret + /* jump into trampoline code */ +1: cmp x18, #HVC_RESET_CPU + b.ne 2f + /* + * Entry point is: + * TRAMPOLINE_VA + * + (__kvm_hyp_reset - (__hyp_idmap_text_start & PAGE_MASK)) + */ + adrp x2, __kvm_hyp_reset + add x2, x2, #:lo12:__kvm_hyp_reset + adrp x3, __hyp_idmap_text_start + add x3, x3, #:lo12:__hyp_idmap_text_start + and x3, x3, PAGE_MASK + sub x2, x2, x3 + ldr x3, =TRAMPOLINE_VA + add x2, x2, x3 + br x2 // no return + +2: /* Default to HVC_CALL_HYP. */ push lr, xzr /*
@@ -999,7 +1021,9 @@ el1_sync: // Guest trapped into EL2 blr lr pop lr, xzr -2: eret + +do_eret: + eret el1_trap: /*
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 267 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)arch/arm/include/asm/kvm_host.h +9
-1@@ -214,6 +214,15 @@ static inline void __cpu_init_hyp_mode(phys_addr_t boot_pgd_ptr,kvm_call_hyp ((void *)hyp_stack_ptr, vector_ptr, pgd_ptr); } +static inline void __cpu_reset_hyp_mode (phys_addr_t boot_pgd_ptr, +phys_addr_t phys_idmap_start) +{ +/* +* TODO +* kvm_call_reset(boot_pgd_ptr, phys_idmap_start); +*/ +} +static inline int kvm_arch_dev_ioctl_check_extension (long ext) {return 0 ;@@ -226,7 +235,6 @@ void kvm_mmu_wp_memory_region(struct kvm *kvm, int slot);struct kvm_vcpu *kvm_mpidr_to_vcpu (struct kvm *kvm,unsigned long mpidr); -static inline void kvm_arch_hardware_disable (void ) {}static inline void kvm_arch_hardware_unsetup (void ) {}static inline void kvm_arch_sync_events (struct kvm *kvm) {}static inline void kvm_arch_vcpu_uninit (struct kvm_vcpu *vcpu) {}arch/arm/include/asm/kvm_mmu.h +1
-0@@ -66,6 +66,7 @@ void kvm_mmu_free_memory_caches(struct kvm_vcpu *vcpu);phys_addr_t kvm_mmu_get_httbr (void );phys_addr_t kvm_mmu_get_boot_httbr (void );phys_addr_t kvm_get_idmap_vector (void ); +phys_addr_t kvm_get_idmap_start (void );int kvm_mmu_init (void );void kvm_clear_hyp_idmap (void );arch/arm/kvm/arm.c +37
-42@@ -16,7 +16,6 @@* Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */ -#include <linux/cpu.h> #include <linux/cpu_pm.h> #include <linux/errno.h> #include <linux/err.h> @@ -61,6 +60,8 @@ static atomic64_t kvm_vmid_gen = ATOMIC64_INIT(1);static u8 kvm_next_vmid;static DEFINE_SPINLOCK (kvm_vmid_lock); +static DEFINE_PER_CPU(unsigned char, kvm_arm_hardware_enabled); +static void kvm_arm_set_running_vcpu (struct kvm_vcpu *vcpu) {BUG_ON (preemptible ());@@ -85,11 +86,6 @@ struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void)return &kvm_arm_running_vcpu; } -int kvm_arch_hardware_enable (void ) -{ -return 0 ; -} -int kvm_arch_vcpu_should_kick (struct kvm_vcpu *vcpu) {return kvm_vcpu_exiting_guest_mode (vcpu) ==IN_GUEST_MODE ;@@ -959,7 +955,7 @@ long kvm_arch_vm_ioctl(struct file *filp, } } -static void cpu_init_hyp_mode (void *dummy) +int kvm_arch_hardware_enable (void ) {phys_addr_t boot_pgd_ptr;phys_addr_t pgd_ptr;@@ -967,6 +963,9 @@ static void cpu_init_hyp_mode(void *dummy)unsigned long stack_page;unsigned long vector_ptr; +if (__hyp_get_vectors () != hyp_default_vectors) +return 0 ; +/* Switch from the HYP stub to our own HYP init vector */ __hyp_set_vectors (kvm_get_idmap_vector ());@@ -979,38 +978,50 @@ static void cpu_init_hyp_mode(void *dummy)__cpu_init_hyp_mode (boot_pgd_ptr, pgd_ptr, hyp_stack_ptr, vector_ptr);kvm_arm_init_debug (); + +return 0 ; } -static int hyp_init_cpu_notify (struct notifier_block *self, -unsigned long action,void *cpu) +void kvm_arch_hardware_disable (void ) { -switch (action) { -case CPU_STARTING : -case CPU_STARTING_FROZEN : -if (__hyp_get_vectors () == hyp_default_vectors) -cpu_init_hyp_mode (NULL ); -break ; - } +phys_addr_t boot_pgd_ptr; +phys_addr_t phys_idmap_start; -return NOTIFY_OK ; -} +if (__hyp_get_vectors () == hyp_default_vectors) +return ; -static struct notifier_block hyp_init_cpu_nb = { - .notifier_call = hyp_init_cpu_notify, -}; + boot_pgd_ptr =kvm_mmu_get_boot_httbr (); + phys_idmap_start =kvm_get_idmap_start (); + +__cpu_reset_hyp_mode (boot_pgd_ptr, phys_idmap_start); +}#ifdef CONFIG_CPU_PM static int hyp_init_cpu_pm_notifier (struct notifier_block *self,unsigned long cmd,void *v) { -if (cmd ==CPU_PM_EXIT && -__hyp_get_vectors () == hyp_default_vectors) { -cpu_init_hyp_mode (NULL ); +switch (cmd) { +case CPU_PM_ENTER : +if (__hyp_get_vectors () != hyp_default_vectors) +__this_cpu_write (kvm_arm_hardware_enabled,1 ); +else +__this_cpu_write (kvm_arm_hardware_enabled,0 ); +/* +* don't call kvm_arch_hardware_disable() in case of +* CPU_PM_ENTER because it does't actually save any state. +*/ + +return NOTIFY_OK ; +case CPU_PM_EXIT : +if (__this_cpu_read (kvm_arm_hardware_enabled)) +kvm_arch_hardware_enable (); +return NOTIFY_OK ; - } -return NOTIFY_DONE ; +default : +return NOTIFY_DONE ; + } } static struct notifier_block hyp_init_cpu_pm_nb = {@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }/* -* Execute the init code on each CPU. -*/ -on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); - -/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();@@ -1186,26 +1192,15 @@ int kvm_arch_init(void *opaque) } } -cpu_notifier_register_begin (); - err =init_hyp_mode ();if (err) goto out_err; - err =__register_cpu_notifier (&hyp_init_cpu_nb); -if (err) { -kvm_err ("Cannot register HYP init CPU notifier (%d)\n" , err); - goto out_err; - } - -cpu_notifier_register_done (); -hyp_cpu_pm_init ();kvm_coproc_table_init ();return 0 ; out_err: -cpu_notifier_register_done ();return err; }arch/arm/kvm/mmu.c +5
-0@@ -1644,6 +1644,11 @@ phys_addr_t kvm_get_idmap_vector(void)return hyp_idmap_vector; } +phys_addr_t kvm_get_idmap_start (void ) +{ +return hyp_idmap_start; +} +int kvm_mmu_init (void ) {int err;arch/arm64/include/asm/kvm_host.h +5
-1@@ -223,6 +223,7 @@ struct kvm_vcpu *kvm_arm_get_running_vcpu(void);struct kvm_vcpu *__percpu *kvm_get_running_vcpus (void );u64 kvm_call_hyp (void *hypfn, ...); +void kvm_call_reset (phys_addr_t boot_pgd_ptr,phys_addr_t phys_idmap_start);void force_vm_exit (const cpumask_t *mask);void kvm_mmu_wp_memory_region (struct kvm *kvm,int slot);@@ -247,7 +248,20 @@ static inline void __cpu_init_hyp_mode(phys_addr_t boot_pgd_ptr, hyp_stack_ptr, vector_ptr); } -static inline void kvm_arch_hardware_disable (void ) {} +static inline void __cpu_reset_hyp_mode (phys_addr_t boot_pgd_ptr, +phys_addr_t phys_idmap_start) +{ +/*
arch/arm64/include/asm/kvm_host.h static inline void __cpu_init_hyp_mode(phys_addr_t boot_pgd_ptr, · line 264
Quoted 10 lines
arch/arm64/include/asm/kvm_host.h +10
-0+* Call reset code, and switch back to stub hyp vectors. +*/ +kvm_call_reset (boot_pgd_ptr, phys_idmap_start); +} + +struct vgic_sr_vectors { +void *save_vgic; +void *restore_vgic; +}; +
Why is this structure being re-introduced? It has been removed in 48f8bd5.
Quoted 25 lines
arch/arm64/include/asm/kvm_host.h +0
-0static inline void kvm_arch_hardware_unsetup (void ) {}static inline void kvm_arch_sync_events (struct kvm *kvm) {}static inline void kvm_arch_vcpu_uninit (struct kvm_vcpu *vcpu) {}arch/arm64/include/asm/kvm_mmu.h +1
-0@@ -98,6 +98,7 @@ void kvm_mmu_free_memory_caches(struct kvm_vcpu *vcpu);phys_addr_t kvm_mmu_get_httbr (void );phys_addr_t kvm_mmu_get_boot_httbr (void );phys_addr_t kvm_get_idmap_vector (void ); +phys_addr_t kvm_get_idmap_start (void );int kvm_mmu_init (void );void kvm_clear_hyp_idmap (void );arch/arm64/include/asm/virt.h +2
-0@@ -58,9 +58,18 @@#define HVC_CALL_FUNC 3 +/* +* HVC_RESET_CPU - Reset cpu in EL2 to initial state.
arch/arm64/include/asm/virt.h line 72
Quoted 10 lines
arch/arm64/include/asm/virt.h +7
-0+* +* @x0: entry address in trampoline code in va +* @x1: identical mapping page table in pa +*/ +#define BOOT_CPU_MODE_EL1 (0xe11)#define BOOT_CPU_MODE_EL2 (0xe12) +#define HVC_RESET_CPU 4 +
Why isn't this next to the comment that document it?
Quoted 13 lines
arch/arm64/include/asm/virt.h +0
-0#ifndef __ASSEMBLY__/* arch/arm64/kvm/hyp-init.S +2
-0@@ -139,6 +139,39 @@ merged: eret ENDPROC(__kvm_hyp_init) + /* + * x0: HYP boot pgd
arch/arm64/kvm/hyp-init.S merged: · line 153
Quoted 10 lines
arch/arm64/kvm/hyp-init.S +10
-0+ * x1: HYP phys_idmap_start + */ +ENTRY(__kvm_hyp_reset) + /* We're in trampoline code in VA, switch back to boot page tables */ + msr ttbr0_el2, x0 + isb + + /* Invalidate the old TLBs */ + tlbi alle2 + dsb sy
I find the location of that TLBI a bit weird. If you want to do something more useful, move it to after the MMU has been disabled.
Quoted 54 lines
arch/arm64/kvm/hyp-init.S +21
-0+ + /* Branch into PA space */ + adr x0, 1f + bfi x1, x0, #0, #PAGE_SHIFT + br x1 + + /* We're now in idmap, disable MMU */ +1: mrs x0, sctlr_el2 + ldr x1, =SCTLR_EL2_FLAGS + bic x0, x0, x1 // Clear SCTL_M and etc + msr sctlr_el2, x0 + isb + + /* Install stub vectors */ + adrp x0, __hyp_stub_vectors + add x0, x0, #:lo12:__hyp_stub_vectors + msr vbar_el2, x0 + + eret +ENDPROC(__kvm_hyp_reset) + .ltorg .popsectionarch/arm64/kvm/hyp.S +11
-3@@ -939,6 +939,11 @@ ENTRY(kvm_call_hyp) ret ENDPROC(kvm_call_hyp) +ENTRY(kvm_call_reset) + hvc #HVC_RESET_CPU + ret +ENDPROC(kvm_call_reset) + .macro invalid_vector label, target .align 2 \label:@@ -982,10 +987,27 @@ el1_sync: // Guest trapped into EL2 cmp x18, #HVC_GET_VECTORS b.ne 1f mrs x0, vbar_el2 - b 2f - -1: /* Default to HVC_CALL_HYP. */ + b do_eret + /* jump into trampoline code */ +1: cmp x18, #HVC_RESET_CPU + b.ne 2f + /* + * Entry point is:
arch/arm64/kvm/hyp.S el1_sync: // Guest trapped into EL2 · line 1006
Quoted 10 lines
arch/arm64/kvm/hyp.S +10
-0+ * TRAMPOLINE_VA + * + (__kvm_hyp_reset - (__hyp_idmap_text_start & PAGE_MASK)) + */ + adrp x2, __kvm_hyp_reset + add x2, x2, #:lo12:__kvm_hyp_reset + adrp x3, __hyp_idmap_text_start + add x3, x3, #:lo12:__hyp_idmap_text_start + and x3, x3, PAGE_MASK + sub x2, x2, x3 + ldr x3, =TRAMPOLINE_VA
Nit: You should be able to do a "mov x3, #TRAMPOLINE_VA and avoid the load from memory.
Quoted 9 lines
arch/arm64/kvm/hyp.S +4
-0+ add x2, x2, x3 + br x2 // no return + +2: /* Default to HVC_CALL_HYP. */ push lr, xzr /*
arch/arm64/kvm/hyp.S el1_sync: // Guest trapped into EL2 · line 1029
Quoted 10 lines
arch/arm64/kvm/hyp.S +3
-1@@ -999,7 +1021,9 @@ el1_sync: // Guest trapped into EL2 blr lr pop lr, xzr -2: eret + +do_eret: + eret el1_trap: /*
Thanks,
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
In reply to Geoff LevandHello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)
[..]
Quoted 5 lines
static struct
notifier_blockhyp_init_cpu_pm_nb= {arch/arm/kvm/arm.c +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }
arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1124
Quoted 10 lines
arch/arm/kvm/arm.c +0
-5/* -* Execute the init code on each CPU. -*/ -on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); - -/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();
With this flow, the cpu_init_hyp_mode() is called only at VM guest creation, but vgic_hyp_init() is called at bootup. On a system with GICv3, it looks like we end up with bogus values from the ICH_VTR_EL2 (to get the number of LRs), because we're not reading it from EL2 anymore.
Whats the best way to fix this?
- Call kvm_arch_hardware_enable() before vgic_hyp_init() and disable later?
- Fold the VGIC init stuff back into hardware_enable()?
- Read the VGIC number of LRs from the hyp stub?
- ..
Regards,
Ashwin.
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 40 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 6 lines
static struct
notifier_blockhyp_init_cpu_pm_nb= {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }
Quoted 15 lines
arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1124
Quoted 9 lines
arch/arm/kvm/arm.c +0
-5/* -* Execute the init code on each CPU. -*/ -on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); - -/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.
..or more likely, that the function to read the ICH_VTR_EL2 is not mapped in EL2, because we deferred the cpu_init_hyp_mode() calls. In any case, this ends up breaking KVM and manifests as a guest console which is stuck.
Quoted 9 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?
- Read the VGIC number of LRs from the hyp stub?
- ..Regards,
Ashwin.
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
In reply to Ashwin ChauguleQuoted 40 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 6 lines
static struct
notifier_blockhyp_init_cpu_pm_nb= {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }
Quoted 15 lines
arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1124
Quoted 9 lines
arch/arm/kvm/arm.c +0
-5/* -* Execute the init code on each CPU. -*/ -on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); - -/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?
None of that works - kvm_arch_hardware_enable() is called once per CPU, while vgic_hyp_init() can only be called once. Also, kvm_arch_hardware_enable() is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.
Quoted 1 line
- Read the VGIC number of LRs from the hyp stub?
That's may UNDEF if called in the wrong context. Also, this defeats the point of stubs, which is just to install the vectors for the hypervisor.
Quoted 1 line
- ..
Yeah, quite.
Geoff, Takahiro?
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Hi All,
Quoted 2 lines · Marc Zyngier
Indeed, this is completely broken (I just reproduced the issue on a model).
There are users out there that want kexec/kdump support. I think we should remove this patch from the series, replace it with a temporary Kconfig conditional that allows only one of KVM or kexec to be enabled, and merge kexec and kdump support in for 4.5. This will satisfy users who need kexec but not KVM, like kexec based bootloaders. Once this KVM hot plug patch is fixed we then merge it, either for 4.5 or 4.6, depending on the timing.
I'll prepare and post a v13 series that does the above.
-Geoff
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 12 lines · Geoff Levand
Quoted 2 lines · Marc Zyngier
Indeed, this is completely broken (I just reproduced the issue on a model).
There are users out there that want kexec/kdump support. I think we should remove this patch from the series, replace it with a temporary
Kconfigconditional that allows only one of KVM or kexec to be enabled, and merge kexec and kdump support in for 4.5. This will satisfy users who need kexec but not KVM, like kexec based bootloaders. Once this KVM hot plug patch is fixed we then merge it, either for 4.5 or 4.6, depending on the timing.
Quoted 1 line
I'll prepare and post a v13 series that does the above.
I'm not really keen on merging this based on a "temporary" Kconfig change with a promise to have it resolved in the future. Given how long this series has been floating around already, I'm not filled with confidence at the prospect of waiting for a fixup patch after it's been merged.
Please address the outstanding review comments before reposting.
Will
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Will,
Quoted 13 lines · Will Deacon
Quoted 12 lines · Geoff Levand
Quoted 2 lines · Marc Zyngier
Indeed, this is completely broken (I just reproduced the issue on a model).
There are users out there that want kexec/kdump support. I think we should remove this patch from the series, replace it with a temporary
Kconfigconditional that allows only one of KVM or kexec to be enabled, and merge kexec and kdump support in for 4.5. This will satisfy users who need kexec but not KVM, like kexec based bootloaders. Once this KVM hot plug patch is fixed we then merge it, either for 4.5 or 4.6, depending on the timing.
Quoted 8 lines
Quoted 1 line
I'll prepare and post a v13 series that does the above.
I'm not really keen on merging this based on a "temporary"
Kconfigchange with a promise to have it resolved in the future. Given how long this series has been floating around already, I'm not filled with confidence at the prospect of waiting for a fixup patch after it's been merged.Please address the outstanding review comments before reposting.
I'm working hard on addressing Marc's comment on kvm cpu hotplug, and so can you please review other *main* part of kexec/kdump patches? Or can I think that 'no comments' is a good sign of your acceptance?
Thanks,
-Takahiro AKASHI
Quoted 2 lines
Will
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
In reply to Marc ZyngierQuoted 57 lines · Marc Zyngier
Quoted 40 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 6 lines
static struct notifier_block hyp_init_cpu_pm_nb = {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }Quoted 15 lines
arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1124
Quoted 9 lines
arch/arm/kvm/arm.c +0
-5/* -* Execute the init code on each CPU. -*/ -on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); - -/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.
Quoted 3 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Sorry for chiming in late. I did run into something similar when I tested some earlier version patches with Akashi San.
The guest bootup just hangs with 4.1 kernel on my LS2085 board which has GICv3. Not sure if this is the same issue. But, my bisect did point to this patch.
I used to think it is my kernel's problem (4.1 sounds a little bit old) since Akashi San can't reproduce this on his Hikey board.
Thanks,
Yang
Quoted 24 lines
Quoted 4 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 1 line
- Read the VGIC number of LRs from the hyp stub?
That's may UNDEF if called in the wrong context. Also, this defeats the point of stubs, which is just to install the vectors for the hypervisor.
Quoted 1 line
- ..
Yeah, quite.
Geoff, Takahiro?
M.
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Yang,
Quoted 58 lines · Yang
Quoted 57 lines · Marc Zyngier
Quoted 55 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 5 lines
static struct notifier_block hyp_init_cpu_pm_nb = {arch/arm/kvm/arm.c +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1124
Quoted 10 lines
arch/arm/kvm/arm.c +0
-5/* -* Execute the init code on each CPU. -*/ -on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); - -/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.
Quoted 8 lines
Quoted 3 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Sorry for chiming in late. I did run into something similar when I tested some earlier version patches with Akashi San.
The guest bootup just hangs with 4.1 kernel on my LS2085 board which has GICv3. Not sure if this is the same issue. But, my bisect did point to this patch.
Can you please try my fixup! patch to determine whether it is the same issue or not?
Thanks
-Takahiro AKASHI
Quoted 31 lines
I used to think it is my kernel's problem (4.1 sounds a little bit old) since Akashi San can't reproduce this on his Hikey board.
Thanks,
YangQuoted 24 lines
Quoted 4 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 1 line
- Read the VGIC number of LRs from the hyp stub?
That's may UNDEF if called in the wrong context. Also, this defeats the point of stubs, which is just to install the vectors for the hypervisor.
Quoted 1 line
- ..
Yeah, quite.
Geoff, Takahiro?
M.
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 74 lines · AKASHI Takahiro
Yang,
Quoted 71 lines · Yang
Quoted 63 lines · Marc Zyngier
Quoted 46 lines · Ashwin Chaugule
Hello,
Quoted 34 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization
at system boot, and has no way to gracefully shutdown a core in terms of
kvm. This prevents, especially, kexec from rebooting the system on a boot
core in EL2.This patch adds a cpu tear-down function and also puts an existing cpu-init
code into a separate function,kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one
stub definition,__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 6 lines
static struct notifier_block hyp_init_cpu_pm_nb = {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }Quoted 15 lines
arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1124
Quoted 9 lines
arch/arm/kvm/arm.c +0
-5/* -* Execute the init code on each CPU. -*/ -on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); - -/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.Quoted 3 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Sorry for chiming in late. I did run into something similar when I tested some earlier version patches with Akashi San.
Quoted 6 lines
Quoted 3 lines
The guest bootup just hangs with 4.1 kernel on my LS2085 board which has GICv3. Not sure if this is the same issue. But,
my bisect did point to this patch.Can you please try my fixup! patch to determine whether it is the same issue or not?
I wish I could help this time, however, unfortunately, my LS2085 board had some hardware failure so I lost the access to it. Once I get a replacement or have it fixed, I will have a try.
Regards,
Yang
Quoted 38 lines
Thanks
-Takahiro AKASHIQuoted 33 lines
I used to think it is my kernel's problem (4.1 sounds a little bit old) since Akashi San can't reproduce this on his
Hikey board.Thanks,
YangQuoted 25 lines
Quoted 5 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 1 line
- Read the VGIC number of LRs from the hyp stub?
That's may UNDEF if called in the wrong context. Also, this defeats the point of stubs, which is just to install the vectors for the hypervisor.
Quoted 1 line
- ..
Yeah, quite.
Geoff, Takahiro?
M.
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
In reply to Marc ZyngierAshwin, Marc,
Quoted 41 lines · Marc Zyngier
Quoted 40 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 6 lines
static struct notifier_block hyp_init_cpu_pm_nb = {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }
Quoted 15 lines
Quoted 10 lines
Review static int init_hyp_mode(void) · line 1124
Quoted 9 lines
(quoted diff) +0
-5/* - * Execute the init code on each CPU. - */ - on_each_cpu(cpu_init_hyp_mode, NULL, 1); - - /* * Init HYP view of VGIC */ err = kvm_vgic_hyp_init();Quoted 5 lines
With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.
Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400, I didn't notice this problem.
Quoted 8 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?
Quoted 5 lines
None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily before kvm_vgic_hyp_init() and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now that cpu_init_hyp_mode() is revived as exactly the same as Marc's original code, the change will not be a big jump.
If kvm_hyp_call() in vgic_v3_probe()/kvm_vgic_hyp_init() is a *problem*, I hope this should work. Actually I confirmed that, with this fixup! patch, we could run a kvm guest and also successfully executed kexec on model w/gic-v3.
My only concern is the following kernel message I saw when kexec shut down the kernel:
(Please note that I was running one kvm quest (pid=961) here.)
===
sh-4.3# ./kexec -d -e
kexec version: 15.11.16.11.06-g41e52e2
arch_process_options:112: command_line: (null)
arch_process_options:114: initrd: (null)
arch_process_options:115: dtb: (null)
arch_process_options:117: port: 0x0
kvm: exiting hardware virtualization
kvm [961]: Unsupported exception type: 6248304 <== this messagekexec_core: Starting new kernel
Disabling non-boot CPUs ...
CPU1: shutdown
CPU2: shutdown
CPU3: shutdown
CPU4: shutdown
CPU5: shutdown
CPU6: shutdown
CPU7: shutdown
Bye!
Booting Linux on physical CPU 0x0
...
===
I don't know whether we can ignore this kind of message or not. Any thoughts?
Thanks,
-Takahiro AKASHI
Quoted 5 lines
Quoted 1 line
- Read the VGIC number of LRs from the hyp stub?
That's may UNDEF if called in the wrong context. Also, this defeats the point of stubs, which is just to install the vectors for the hypervisor.
Quoted 8 lines
Quoted 1 line
- ..
Yeah, quite.
Geoff, Takahiro?
M.
----8<----
From 66ca3baedf45c78c81a76ea77ddd6ace49550ab6 Mon Sep 17 00:00:00 2001
From: AKASHI Takahiro <[email protected]>
Date: Fri, 11 Dec 2015 13:43:35 +0900
Subject: [PATCH 7/7] fixup! arm64: kvm: allows kvm cpu hotplug
--- arch/arm/kvm/arm.c | 37 +++++++++++++++++++++++++++---------- 1 file changed, 27 insertions(+), 10 deletions(-)
arch/arm/kvm/arm.c +27 -10
@@ -950,7 +950,7 @@ long kvm_arch_vm_ioctl(struct file *filp, } } -int kvm_arch_hardware_enable (void ) +static void cpu_init_hyp_mode (void *dummy) {phys_addr_t boot_pgd_ptr;phys_addr_t pgd_ptr;
@@ -958,9 +958,6 @@ int kvm_arch_hardware_enable(void)unsigned long stack_page;unsigned long vector_ptr; -if (__hyp_get_vectors () != hyp_default_vectors) -return 0 ; -/* Switch from the HYP stub to our own HYP init vector */ __hyp_set_vectors (kvm_get_idmap_vector ());
@@ -973,24 +970,35 @@ int kvm_arch_hardware_enable(void)__cpu_init_hyp_mode (boot_pgd_ptr, pgd_ptr, hyp_stack_ptr, vector_ptr);kvm_arm_init_debug (); - -return 0 ; } -void kvm_arch_hardware_disable (void ) +static void cpu_reset_hyp_mode (void *dummy) {phys_addr_t boot_pgd_ptr;phys_addr_t phys_idmap_start; -if (__hyp_get_vectors () == hyp_default_vectors) -return ; - boot_pgd_ptr =kvm_mmu_get_boot_httbr (); phys_idmap_start =kvm_get_idmap_start ();__cpu_reset_hyp_mode (boot_pgd_ptr, phys_idmap_start); } +int kvm_arch_hardware_enable (void ) +{ +if (__hyp_get_vectors () == hyp_default_vectors) +cpu_init_hyp_mode (NULL ); + +return 0 ; +} + +void kvm_arch_hardware_disable (void ) +{ +if (__hyp_get_vectors () == hyp_default_vectors) +return ; + +cpu_reset_hyp_mode (NULL ); +} +#ifdef CONFIG_CPU_PM static int hyp_init_cpu_pm_notifier(struct notifier_block *self, unsigned long cmd,
@@ -1114,6 +1122,12 @@ static int init_hyp_mode(void) }/* +* Execute the init code on each CPU. +* Only needed to execute kvm_hyp_call() during *_hyp_init(). +*/ +on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); + +/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();
@@ -1127,6 +1141,8 @@ static int init_hyp_mode(void)if (err) goto out_free_context; +on_each_cpu (cpu_reset_hyp_mode,NULL ,1 ); +#ifndef CONFIG_HOTPLUG_CPU free_boot_hyp_pgd ();#endif
@@ -1137,6 +1153,7 @@ static int init_hyp_mode(void)return 0 ; out_free_context: +on_each_cpu (cpu_reset_hyp_mode,NULL ,1 );free_percpu (kvm_host_cpu_state); out_free_mappings:free_hyp_pgds ();
Signature · 1 line
1.7.9.5
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 86 lines · AKASHI Takahiro
Ashwin, Marc,
Quoted 7 lines · Marc Zyngier
Quoted 6 lines · Ashwin Chaugule
Hello,
Quoted 3 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific
initialization
Quoted 1 line
Quoted 1 line
Quoted 1 line
at system boot, and has no way to gracefully shutdown a core in terms
of
Quoted 1 line
Quoted 1 line
Quoted 1 line
kvm. This prevents, especially, kexec from rebooting the system on a
boot
Quoted 3 lines
Quoted 3 lines
Quoted 3 lines
core in EL2.
This patch adds a cpu tear-down function and also puts an existing
cpu-init
Quoted 5 lines
Quoted 5 lines
Quoted 5 lines
code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64,
one
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79++++++++++++++++++---------------------
Quoted 16 lines
Quoted 16 lines
Quoted 7 lines
arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 6 lines
static struct notifier_block hyp_init_cpu_pm_nb = {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }Quoted 15 lines
Quoted 10 lines
Review static int init_hyp_mode(void) · line 1124
Quoted 9 lines
(quoted diff) +0
-5/* - * Execute the init code on each CPU. - */ - on_each_cpu(cpu_init_hyp_mode, NULL, 1); - - /* * Init HYP view of VGIC */ err = kvm_vgic_hyp_init();Quoted 5 lines
With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400,
I didn't notice this problem.Quoted 6 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 2 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disablelater?
Quoted 7 lines
Quoted 1 line
- Fold the VGIC init stuff back into
hardware_enable()?None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.
Quoted 8 lines
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily
beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.
I have tested this patch and verified the fix on our hardware with GICv3 without any issue.
Quoted 154 lines
If
kvm_hyp_call()in vgic_v3_probe()/kvm_vgic_hyp_init() is a *problem*, I hope this should work. Actually I confirmed that, with this fixup! patch,
we could run a kvm guest and also successfully executed kexec on model w/gic-v3.My only concern is the following kernel message I saw when kexec shut down the kernel:
(Please note that I was running one kvm quest (pid=961) here.)===
sh-4.3# ./kexec -d -e
kexec version: 15.11.16.11.06-g41e52e2
arch_process_options:112:command_line: (null)
arch_process_options:114: initrd: (null)
arch_process_options:115: dtb: (null)
arch_process_options:117: port:0x0
kvm: exiting hardware virtualization
kvm [961]: Unsupported exception type: 6248304 <== this messagekexec_core: Starting new kernel
Disabling non-boot CPUs ...
CPU1: shutdown
CPU2: shutdown
CPU3: shutdown
CPU4: shutdown
CPU5: shutdown
CPU6: shutdown
CPU7: shutdown
Bye!
Booting Linux on physical CPU0x0
...
===I don't know whether we can ignore this kind of message or not. Any thoughts?
Thanks,
-Takahiro AKASHIQuoted 5 lines
Quoted 1 line
- Read the VGIC number of LRs from the hyp stub?
That's may UNDEF if called in the wrong context. Also, this defeats the point of stubs, which is just to install the vectors for the hypervisor.
Quoted 8 lines
Quoted 1 line
- ..
Yeah, quite.
Geoff, Takahiro?
M.----8<----
From66ca3baedf45c78c81a76ea77ddd6ace49550ab6Mon Sep 17 00:00:00 2001
From: AKASHI Takahiro <[email protected]>
Date: Fri, 11 Dec 2015 13:43:35 +0900
Subject: [PATCH 7/7] fixup! arm64: kvm: allows kvm cpu hotplug--- arch/arm/kvm/arm.c | 37 +++++++++++++++++++++++++++---------- 1 file changed, 27 insertions(+), 10 deletions(-)arch/arm/kvm/arm.c +18
-10@@ -950,7 +950,7 @@ long kvm_arch_vm_ioctl(struct file *filp, } } -int kvm_arch_hardware_enable (void ) +static void cpu_init_hyp_mode (void *dummy) {phys_addr_t boot_pgd_ptr;phys_addr_t pgd_ptr;@@ -958,9 +958,6 @@ int kvm_arch_hardware_enable(void)unsigned long stack_page;unsigned long vector_ptr; -if (__hyp_get_vectors () != hyp_default_vectors) -return 0 ; -/* Switch from the HYP stub to our own HYP init vector */ __hyp_set_vectors (kvm_get_idmap_vector ());@@ -973,24 +970,35 @@ int kvm_arch_hardware_enable(void)__cpu_init_hyp_mode (boot_pgd_ptr, pgd_ptr, hyp_stack_ptr, vector_ptr);kvm_arm_init_debug (); - -return 0 ; } -void kvm_arch_hardware_disable (void ) +static void cpu_reset_hyp_mode (void *dummy) {phys_addr_t boot_pgd_ptr;phys_addr_t phys_idmap_start; -if (__hyp_get_vectors () == hyp_default_vectors) -return ; - boot_pgd_ptr =kvm_mmu_get_boot_httbr (); phys_idmap_start =kvm_get_idmap_start ();__cpu_reset_hyp_mode (boot_pgd_ptr, phys_idmap_start); } +int kvm_arch_hardware_enable (void ) +{ +if (__hyp_get_vectors () == hyp_default_vectors) +cpu_init_hyp_mode (NULL ); + +return 0 ; +} + +void kvm_arch_hardware_disable (void ) +{ +if (__hyp_get_vectors () == hyp_default_vectors) +return ; + +cpu_reset_hyp_mode (NULL ); +} +#ifdef CONFIG_CPU_PM static int hyp_init_cpu_pm_notifier(struct notifier_block *self,unsigned long cmd,
arch/arm/kvm/arm.c +8
-0@@ -1114,6 +1122,12 @@ static int init_hyp_mode(void) }/* +* Execute the init code on each CPU. +* Only needed to execute kvm_hyp_call() during *_hyp_init(). +*/ +on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); + +/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();@@ -1127,6 +1141,8 @@ static int init_hyp_mode(void)if (err) goto out_free_context; +on_each_cpu (cpu_reset_hyp_mode,NULL ,1 ); +#ifndef CONFIG_HOTPLUG_CPU free_boot_hyp_pgd ();#endif
arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1159
Quoted 7 lines
arch/arm/kvm/arm.c +1
-0@@ -1137,6 +1153,7 @@ static int init_hyp_mode(void)return 0 ; out_free_context: +on_each_cpu (cpu_reset_hyp_mode,NULL ,1 );free_percpu (kvm_host_cpu_state); out_free_mappings:free_hyp_pgds ();
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
In reply to AKASHI TakahiroQuoted 60 lines · AKASHI Takahiro
Ashwin, Marc,
Quoted 41 lines · Marc Zyngier
Quoted 40 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 6 lines
static struct notifier_block hyp_init_cpu_pm_nb = {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }Quoted 15 lines
Quoted 10 lines
Review static int init_hyp_mode(void) · line 1124
Quoted 9 lines
(quoted diff) +0
-5/* - * Execute the init code on each CPU. - */ - on_each_cpu(cpu_init_hyp_mode, NULL, 1); - - /* * Init HYP view of VGIC */ err = kvm_vgic_hyp_init();Quoted 5 lines
With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.
Quoted 3 lines
Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400, I didn't notice this problem.
Because GIC-400 is a GICv2 implementation, which is entirely MMIO based. GICv3 uses some system registers that are only available at EL2, and KVM needs some information contained in these registers before being able to get initialized.
Quoted 14 lines
Quoted 8 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?Quoted 5 lines
None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.
Quoted 7 lines
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.
This seems quite complicated:
- init EL2 on all CPUs
- do some initialization
- tear all CPUs EL2 down
- let KVM drive the vectors being set or not
My questions are: why do we need to do this on *all* cpus? Can't that work on a single one?
Also, the simple fact that we were able to get some junk value is a sign that something is amiss. I'd expect a splat of some sort, because we now have a possibility of doing things in the wrong context.
Quoted 9 lines
If
kvm_hyp_call()in vgic_v3_probe()/kvm_vgic_hyp_init() is a *problem*, I hope this should work. Actually I confirmed that, with this fixup! patch, we could run a kvm guest and also successfully executed kexec on model w/gic-v3.My only concern is the following kernel message I saw when kexec shut down the kernel:
(Please note that I was running one kvm quest (pid=961) here.)
Quoted 9 lines
===
sh-4.3# ./kexec -d -e
kexec version: 15.11.16.11.06-g41e52e2
arch_process_options:112:command_line: (null)
arch_process_options:114: initrd: (null)
arch_process_options:115: dtb: (null)
arch_process_options:117: port:0x0
kvm: exiting hardware virtualization
kvm [961]: Unsupported exception type: 6248304 <== this message
That makes me feel very uncomfortable. It looks like we've exited a guest with some horrible value in X0. How is that even possible?
This deserves to be investigated.
Thanks,
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 84 lines · Marc Zyngier
Quoted 56 lines · AKASHI Takahiro
Ashwin, Marc,
Quoted 39 lines · Marc Zyngier
Quoted 38 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 5 lines
static struct notifier_block hyp_init_cpu_pm_nb = {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }Quoted 14 lines
Quoted 9 lines
Review static int init_hyp_mode(void) · line 1124
Quoted 9 lines
(quoted diff) +0
-5/* - * Execute the init code on each CPU. - */ - on_each_cpu(cpu_init_hyp_mode, NULL, 1); - - /* * Init HYP view of VGIC */ err = kvm_vgic_hyp_init();Quoted 5 lines
With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.Quoted 3 lines
Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400, I didn't notice this problem.Because GIC-400 is a GICv2 implementation, which is entirely MMIO based. GICv3 uses some system registers that are only available at EL2, and KVM needs some information contained in these registers before being able to get initialized.
Quoted 12 lines
Quoted 4 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 8 lines
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 7 lines
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.
Quoted 9 lines
This seems quite complicated:
- init EL2 on all CPUs
- do some initialization
- tear all CPUs EL2 down
- let KVM drive the vectors being set or notMy questions are: why do we need to do this on *all* cpus? Can't that work on a single one?
Single CPU EL2 initialization should be fine as long as no kernel preemption happens in between init EL2 and kvm_vgic_hyp_init() execution. The function init_hyp_mode() is called by do_basic_setup() with preemption enabled.
I don't have deeper knowledge of how scheduler is handled during the kernel boot time, but initializing all CPUs definitely helps if preemption happens before reading ICH_VTR_EL2 register and after kvm_vgic_hyp_init().
Quoted 21 lines
Also, the simple fact that we were able to get some junk value is a sign that something is amiss. I'd expect a splat of some sort, because we now have a possibility of doing things in the wrong context.
Quoted 8 lines
If
kvm_hyp_call()in vgic_v3_probe()/kvm_vgic_hyp_init() is a *problem*, I hope this should work. Actually I confirmed that, with this fixup! patch, we could run a kvm guest and also successfully executed kexec on model w/gic-v3.My only concern is the following kernel message I saw when kexec shut down the kernel:
(Please note that I was running one kvm quest (pid=961) here.)Quoted 9 lines
===
sh-4.3# ./kexec -d -e
kexec version: 15.11.16.11.06-g41e52e2
arch_process_options:112:command_line: (null)
arch_process_options:114: initrd: (null)
arch_process_options:115: dtb: (null)
arch_process_options:117: port:0x0
kvm: exiting hardware virtualization
kvm [961]: Unsupported exception type: 6248304 <== this message
Quoted 8 lines
That makes me feel very uncomfortable. It looks like we've exited a guest with some horrible value in X0. How is that even possible?
This deserves to be investigated.
Thanks,
M.
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 93 lines · Shanker Donthineni
Quoted 90 lines · Marc Zyngier
Quoted 56 lines · AKASHI Takahiro
Ashwin, Marc,
Quoted 39 lines · Marc Zyngier
Quoted 38 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 5 lines
static struct notifier_block hyp_init_cpu_pm_nb = {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }Quoted 14 lines
Quoted 9 lines
Review static int init_hyp_mode(void) · line 1124
Quoted 9 lines
(quoted diff) +0
-5/* - * Execute the init code on each CPU. - */ - on_each_cpu(cpu_init_hyp_mode, NULL, 1); - - /* * Init HYP view of VGIC */ err = kvm_vgic_hyp_init();Quoted 5 lines
With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.Quoted 3 lines
Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400, I didn't notice this problem.Because GIC-400 is a GICv2 implementation, which is entirely MMIO based. GICv3 uses some system registers that are only available at EL2, and KVM needs some information contained in these registers before being able to get initialized.
Quoted 12 lines
Quoted 4 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 8 lines
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 7 lines
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.This seems quite complicated:
- init EL2 on all CPUs
- do some initialization
- tear all CPUs EL2 down
- let KVM drive the vectors being set or not
Quoted 8 lines
Quoted 3 lines
My questions are: why do we need to do this on *all* cpus? Can't that work on a single one?
Single CPU EL2 initialization should be fine as long as no kernel preemption happens in between init EL2 and
kvm_vgic_hyp_init()execution. The functioninit_hyp_mode()is called bydo_basic_setup()with preemption enabled.
Indeed. So far, we never needed this since we were executing this code with interrupts disabled.
Quoted 4 lines
I don't have deeper knowledge of how scheduler is handled during the
kernel boot time, but initializing all CPUs definitely helps if
preemption happens before readingICH_VTR_EL2register and afterkvm_vgic_hyp_init().
What is wrong with wrapping the critical path with preempt_{enabled,disabled}?
Thanks,
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 102 lines · Marc Zyngier
Quoted 91 lines · Shanker Donthineni
Quoted 89 lines · Marc Zyngier
Quoted 55 lines · AKASHI Takahiro
Ashwin, Marc,
Quoted 38 lines · Marc Zyngier
Quoted 37 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 4 lines
static struct notifier_block hyp_init_cpu_pm_nb = {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }Quoted 14 lines
Quoted 9 lines
Review static int init_hyp_mode(void) · line 1124
Quoted 9 lines
(quoted diff) +0
-5/* - * Execute the init code on each CPU. - */ - on_each_cpu(cpu_init_hyp_mode, NULL, 1); - - /* * Init HYP view of VGIC */ err = kvm_vgic_hyp_init();Quoted 5 lines
With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.Quoted 3 lines
Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400, I didn't notice this problem.Because GIC-400 is a GICv2 implementation, which is entirely MMIO based. GICv3 uses some system registers that are only available at EL2, and KVM needs some information contained in these registers before being able to get initialized.
Quoted 12 lines
Quoted 4 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 8 lines
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 7 lines
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.This seems quite complicated:
- init EL2 on all CPUs
- do some initialization
- tear all CPUs EL2 down
- let KVM drive the vectors being set or notQuoted 7 lines
Quoted 3 lines
My questions are: why do we need to do this on *all* cpus? Can't that work on a single one?
Single CPU EL2 initialization should be fine as long as no kernel preemption happens in between init EL2 and
kvm_vgic_hyp_init()execution. The functioninit_hyp_mode()is called bydo_basic_setup()with preemption enabled.Indeed. So far, we never needed this since we were executing this code with interrupts disabled.
Quoted 6 lines
Quoted 4 lines
I don't have deeper knowledge of how scheduler is handled during the
kernel boot time, but initializing all CPUs definitely helps if
preemption happens before readingICH_VTR_EL2register and afterkvm_vgic_hyp_init().What is wrong with wrapping the critical path with preempt_{enabled,disabled}?
I agree with you nothing wrong implementing your suggestion. It will fix the problem.
Quoted 3 lines
Thanks,
M.
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
In reply to Marc ZyngierQuoted 5 lines · Marc Zyngier
Quoted 4 lines · AKASHI Takahiro
Quoted 3 lines · Marc Zyngier
Quoted 2 lines · Ashwin Chaugule
Quoted 1 line · Geoff Levand
From: AKASHI Takahiro <[email protected]>
[..]
Quoted 13 lines
Quoted 6 lines
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.This seems quite complicated:
- init EL2 on all CPUs
- do some initialization
- tear all CPUs EL2 down
- let KVM drive the vectors being set or not
Quoted 6 lines
My questions are: why do we need to do this on *all* cpus? Can't that work on a single one?
Also, the simple fact that we were able to get some junk value is a sign that something is amiss. I'd expect a splat of some sort, because we now have a possibility of doing things in the wrong context.
Yea. I had the same expectation too. Probably some leftover state from the hyp stub at bootup kept us lucky.
Quoted 20 lines
Quoted 10 lines
If
kvm_hyp_call()in vgic_v3_probe()/kvm_vgic_hyp_init() is a *problem*, I hope this should work. Actually I confirmed that, with this fixup! patch, we could run a kvm guest and also successfully executed kexec on model w/gic-v3.My only concern is the following kernel message I saw when kexec shut down the kernel:
(Please note that I was running one kvm quest (pid=961) here.)Quoted 9 lines
===
sh-4.3# ./kexec -d -e
kexec version: 15.11.16.11.06-g41e52e2
arch_process_options:112:command_line: (null)
arch_process_options:114: initrd: (null)
arch_process_options:115: dtb: (null)
arch_process_options:117: port:0x0
kvm: exiting hardware virtualization
kvm [961]: Unsupported exception type: 6248304 <== this message
Quoted 4 lines
That makes me feel very uncomfortable. It looks like we've exited a guest with some horrible value in X0. How is that even possible?
This deserves to be investigated.
Did this exception trigger as a result of the fixup? For giggles, what happens if you dont tear down the EL2 setup after reading the number of LRs? Are we missing another site where we need to setup and teardown?
Ashwin
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
In reply to Marc ZyngierMarc,
Quoted 61 lines · Marc Zyngier
Quoted 60 lines · AKASHI Takahiro
Ashwin, Marc,
Quoted 56 lines · Marc Zyngier
Quoted 55 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 5 lines
static struct notifier_block hyp_init_cpu_pm_nb = {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }Review static int init_hyp_mode(void) · line 1124
Quoted 10 lines
(quoted diff) +0
-5/* - * Execute the init code on each CPU. - */ - on_each_cpu(cpu_init_hyp_mode, NULL, 1); - - /* * Init HYP view of VGIC */ err = kvm_vgic_hyp_init();With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.
Quoted 8 lines
Quoted 3 lines
Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400, I didn't notice this problem.Because GIC-400 is a GICv2 implementation, which is entirely MMIO based. GICv3 uses some system registers that are only available at EL2, and KVM needs some information contained in these registers before being able to get initialized.
I see.
Quoted 22 lines
Quoted 14 lines
Quoted 8 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?Quoted 5 lines
None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 7 lines
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.
Quoted 8 lines
This seems quite complicated:
- init EL2 on all CPUs
- do some initialization
- tear all CPUs EL2 down
- let KVM drive the vectors being set or notMy questions are: why do we need to do this on *all* cpus? Can't that work on a single one?
I did initialize all the cpus partly because using preempt_enable/disable looked a bit ugly and partly because we may, in the future, do additional per-cpu initialization in kvm_vgic_hyp_init() and/or kvm_timer_hyp_init(). But if you're comfortable with preempt_*() stuff, I don' care.
Quoted 23 lines
Also, the simple fact that we were able to get some junk value is a sign that something is amiss. I'd expect a splat of some sort, because we now have a possibility of doing things in the wrong context.
Quoted 9 lines
If
kvm_hyp_call()in vgic_v3_probe()/kvm_vgic_hyp_init() is a *problem*, I hope this should work. Actually I confirmed that, with this fixup! patch, we could run a kvm guest and also successfully executed kexec on model w/gic-v3.My only concern is the following kernel message I saw when kexec shut down the kernel:
(Please note that I was running one kvm quest (pid=961) here.)Quoted 9 lines
===
sh-4.3# ./kexec -d -e
kexec version: 15.11.16.11.06-g41e52e2
arch_process_options:112:command_line: (null)
arch_process_options:114: initrd: (null)
arch_process_options:115: dtb: (null)
arch_process_options:117: port:0x0
kvm: exiting hardware virtualization
kvm [961]: Unsupported exception type: 6248304 <== this message
Quoted 4 lines
That makes me feel very uncomfortable. It looks like we've exited a guest with some horrible value in X0. How is that even possible?
This deserves to be investigated.
I guess the problem is that cpu tear-down function is called even if a kvm guest is still running in kvm_arch_vcpu_ioctl_run().
So adding a check whether cpu has been initialized or not in every iteration of kvm_arch_vcpu_ioctl_run() will, if necessary, terminate a guest safely without entering a guest mode. Since this check is done while interrupt is disabled, it won't interfere with kvm_arch_hardware_disable() called via IPI.
See the attached fixup patch.
Again, I verified the code on model.
Thanks,
-Takahiro AKASHI
Quoted 4 lines
Thanks,
M.
----8<----
From 77f273ba5e0c3dfcf75a5a8d1da8035cc390250c Mon Sep 17 00:00:00 2001
From: AKASHI Takahiro <[email protected]>
Date: Fri, 11 Dec 2015 13:43:35 +0900
Subject: [PATCH] fixup! arm64: kvm: allows kvm cpu hotplug
--- arch/arm/kvm/arm.c | 45 ++++++++++++++++++++++++++++++++++----------- 1 file changed, 34 insertions(+), 11 deletions(-)
arch/arm/kvm/arm.c +34 -11
@@ -573,7 +573,11 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)/* * Re-check atomic conditions */ -if (signal_pending (current)) { +if (__hyp_get_vectors () == hyp_default_vectors) { +/* cpu has been torn down */ + ret = -ENOEXEC ; + run->exit_reason =KVM_EXIT_SHUTDOWN ; + }else if (signal_pending (current)) { ret = -EINTR ; run->exit_reason =KVM_EXIT_INTR ; }
@@ -950,7 +954,7 @@ long kvm_arch_vm_ioctl(struct file *filp, } } -int kvm_arch_hardware_enable (void ) +static void cpu_init_hyp_mode (void ) {phys_addr_t boot_pgd_ptr;phys_addr_t pgd_ptr;
@@ -958,9 +962,6 @@ int kvm_arch_hardware_enable(void)unsigned long stack_page;unsigned long vector_ptr; -if (__hyp_get_vectors () != hyp_default_vectors) -return 0 ; -/* Switch from the HYP stub to our own HYP init vector */ __hyp_set_vectors (kvm_get_idmap_vector ());
@@ -973,24 +974,35 @@ int kvm_arch_hardware_enable(void)__cpu_init_hyp_mode (boot_pgd_ptr, pgd_ptr, hyp_stack_ptr, vector_ptr);kvm_arm_init_debug (); - -return 0 ; } -void kvm_arch_hardware_disable (void ) +static void cpu_reset_hyp_mode (void ) {phys_addr_t boot_pgd_ptr;phys_addr_t phys_idmap_start; -if (__hyp_get_vectors () == hyp_default_vectors) -return ; - boot_pgd_ptr =kvm_mmu_get_boot_httbr (); phys_idmap_start =kvm_get_idmap_start ();__cpu_reset_hyp_mode (boot_pgd_ptr, phys_idmap_start); } +int kvm_arch_hardware_enable (void ) +{ +if (__hyp_get_vectors () == hyp_default_vectors) +cpu_init_hyp_mode (); + +return 0 ; +} + +void kvm_arch_hardware_disable (void ) +{ +if (__hyp_get_vectors () == hyp_default_vectors) +return ; + +cpu_reset_hyp_mode (); +} +#ifdef CONFIG_CPU_PM static int hyp_init_cpu_pm_notifier(struct notifier_block *self, unsigned long cmd,
@@ -1114,9 +1126,20 @@ static int init_hyp_mode(void) }/* +* Init this CPU temporarily to execute kvm_hyp_call() +* during kvm_vgic_hyp_init(). +*/ +preempt_disable (); +cpu_init_hyp_mode (); + +/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init (); + +cpu_reset_hyp_mode (); +preempt_enable (); +if (err) goto out_free_context;
Signature · 1 line
1.7.9.5
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 173 lines · AKASHI Takahiro
Marc,
Quoted 61 lines · Marc Zyngier
Quoted 60 lines · AKASHI Takahiro
Ashwin, Marc,
Quoted 56 lines · Marc Zyngier
Quoted 55 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 5 lines
static struct notifier_block hyp_init_cpu_pm_nb = {arch/arm/kvm/arm.c +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1124
Quoted 10 lines
arch/arm/kvm/arm.c +0
-5/* -* Execute the init code on each CPU. -*/ -on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); - -/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.Quoted 8 lines
Quoted 3 lines
Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400, I didn't notice this problem.Because GIC-400 is a GICv2 implementation, which is entirely MMIO based. GICv3 uses some system registers that are only available at EL2, and KVM needs some information contained in these registers before being able to get initialized.
I see.
Quoted 22 lines
Quoted 14 lines
Quoted 8 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?Quoted 5 lines
None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 7 lines
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.Quoted 8 lines
This seems quite complicated:
- init EL2 on all CPUs
- do some initialization
- tear all CPUs EL2 down
- let KVM drive the vectors being set or notMy questions are: why do we need to do this on *all* cpus? Can't that work on a single one?
I did initialize all the cpus partly because using preempt_enable/disable looked a bit ugly and partly because we may, in the future, do additional per-cpu initialization in
kvm_vgic_hyp_init()and/orkvm_timer_hyp_init(). But if you're comfortable with preempt_*() stuff, I don' care.Quoted 23 lines
Also, the simple fact that we were able to get some junk value is a sign that something is amiss. I'd expect a splat of some sort, because we now have a possibility of doing things in the wrong context.
Quoted 9 lines
If
kvm_hyp_call()in vgic_v3_probe()/kvm_vgic_hyp_init() is a *problem*, I hope this should work. Actually I confirmed that, with this fixup! patch, we could run a kvm guest and also successfully executed kexec on model w/gic-v3.My only concern is the following kernel message I saw when kexec shut down the kernel:
(Please note that I was running one kvm quest (pid=961) here.)Quoted 9 lines
===
sh-4.3# ./kexec -d -e
kexec version: 15.11.16.11.06-g41e52e2
arch_process_options:112:command_line: (null)
arch_process_options:114: initrd: (null)
arch_process_options:115: dtb: (null)
arch_process_options:117: port:0x0
kvm: exiting hardware virtualization
kvm [961]: Unsupported exception type: 6248304 <== this messageQuoted 4 lines
That makes me feel very uncomfortable. It looks like we've exited a guest with some horrible value in X0. How is that even possible?
This deserves to be investigated.
I guess the problem is that cpu tear-down function is called even if a kvm guest is still running in
kvm_arch_vcpu_ioctl_run().
So adding a check whether cpu has been initialized or not in every iteration ofkvm_arch_vcpu_ioctl_run()will, if necessary, terminate a guest safely without entering a guest mode. Since this check is done while interrupt is disabled, it won't interfere withkvm_arch_hardware_disable()called via IPI.
See the attached fixup patch.Again, I verified the code on model.
Thanks,
-Takahiro AKASHIQuoted 4 lines
Thanks,
M.----8<----
From77f273ba5e0c3dfcf75a5a8d1da8035cc390250cMon Sep 17 00:00:00 2001
From: AKASHI Takahiro <[email protected]>
Date: Fri, 11 Dec 2015 13:43:35 +0900
Subject: [PATCH] fixup! arm64: kvm: allows kvm cpu hotplug--- arch/arm/kvm/arm.c | 45 ++++++++++++++++++++++++++++++++++----------- 1 file changed, 34 insertions(+), 11 deletions(-)arch/arm/kvm/arm.c +0
-0
arch/arm/kvm/arm.c int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run) · line 579
Quoted 8 lines
arch/arm/kvm/arm.c +4
-1@@ -573,7 +573,11 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)/* * Re-check atomic conditions */ -if (signal_pending (current)) { +if (__hyp_get_vectors () == hyp_default_vectors) { +/* cpu has been torn down */ + ret = -ENOEXEC ; + run->exit_reason =KVM_EXIT_SHUTDOWN ;
That feels completely overkill (and very slow). Why don't you maintain a per-cpu variable containing the CPU states, which will avoid calling __hyp_get_vectors() all the time? You should be able to reuse that construct everywhere.
Also, I'm not sure about KVM_EXIT_SHUTDOWN. This looks very x86 specific (called on triple fault). KVM_EXIT_FAIL_ENTRY looks more appropriate, and the hardware_entry_failure_reason field should be populated (and documented).
Thanks,
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 180 lines · Marc Zyngier
Quoted 179 lines · AKASHI Takahiro
Marc,
Quoted 61 lines · Marc Zyngier
Quoted 60 lines · AKASHI Takahiro
Ashwin, Marc,
Quoted 56 lines · Marc Zyngier
Quoted 55 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 5 lines
static struct notifier_block hyp_init_cpu_pm_nb = {arch/arm/kvm/arm.c +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1124
Quoted 10 lines
arch/arm/kvm/arm.c +0
-5/* -* Execute the init code on each CPU. -*/ -on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); - -/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.Quoted 8 lines
Quoted 3 lines
Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400, I didn't notice this problem.Because GIC-400 is a GICv2 implementation, which is entirely MMIO based. GICv3 uses some system registers that are only available at EL2, and KVM needs some information contained in these registers before being able to get initialized.
I see.
Quoted 22 lines
Quoted 14 lines
Quoted 8 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?Quoted 5 lines
None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 7 lines
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.Quoted 8 lines
This seems quite complicated:
- init EL2 on all CPUs
- do some initialization
- tear all CPUs EL2 down
- let KVM drive the vectors being set or notMy questions are: why do we need to do this on *all* cpus? Can't that work on a single one?
I did initialize all the cpus partly because using preempt_enable/disable looked a bit ugly and partly because we may, in the future, do additional per-cpu initialization in
kvm_vgic_hyp_init()and/orkvm_timer_hyp_init(). But if you're comfortable with preempt_*() stuff, I don' care.Quoted 23 lines
Also, the simple fact that we were able to get some junk value is a sign that something is amiss. I'd expect a splat of some sort, because we now have a possibility of doing things in the wrong context.
Quoted 9 lines
If
kvm_hyp_call()in vgic_v3_probe()/kvm_vgic_hyp_init() is a *problem*, I hope this should work. Actually I confirmed that, with this fixup! patch, we could run a kvm guest and also successfully executed kexec on model w/gic-v3.My only concern is the following kernel message I saw when kexec shut down the kernel:
(Please note that I was running one kvm quest (pid=961) here.)Quoted 9 lines
===
sh-4.3# ./kexec -d -e
kexec version: 15.11.16.11.06-g41e52e2
arch_process_options:112:command_line: (null)
arch_process_options:114: initrd: (null)
arch_process_options:115: dtb: (null)
arch_process_options:117: port:0x0
kvm: exiting hardware virtualization
kvm [961]: Unsupported exception type: 6248304 <== this messageQuoted 4 lines
That makes me feel very uncomfortable. It looks like we've exited a guest with some horrible value in X0. How is that even possible?
This deserves to be investigated.
I guess the problem is that cpu tear-down function is called even if a kvm guest is still running in
kvm_arch_vcpu_ioctl_run().
So adding a check whether cpu has been initialized or not in every iteration ofkvm_arch_vcpu_ioctl_run()will, if necessary, terminate a guest safely without entering a guest mode. Since this check is done while interrupt is disabled, it won't interfere withkvm_arch_hardware_disable()called via IPI.
See the attached fixup patch.Again, I verified the code on model.
Thanks,
-Takahiro AKASHIQuoted 4 lines
Thanks,
M.----8<----
From77f273ba5e0c3dfcf75a5a8d1da8035cc390250cMon Sep 17 00:00:00 2001
From: AKASHI Takahiro <[email protected]>
Date: Fri, 11 Dec 2015 13:43:35 +0900
Subject: [PATCH] fixup! arm64: kvm: allows kvm cpu hotplug--- arch/arm/kvm/arm.c | 45 ++++++++++++++++++++++++++++++++++----------- 1 file changed, 34 insertions(+), 11 deletions(-)arch/arm/kvm/arm.c +2
-1@@ -573,7 +573,11 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)/* * Re-check atomic conditions */ -if (signal_pending (current)) { +if (__hyp_get_vectors () == hyp_default_vectors) { +/* cpu has been torn down */
Quoted 8 lines
arch/arm/kvm/arm.c int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run) · line 579
Quoted 2 lines
arch/arm/kvm/arm.c +2
-0+ ret = -ENOEXEC ; + run->exit_reason =KVM_EXIT_SHUTDOWN ;That feels completely overkill (and very slow). Why don't you maintain a per-cpu variable containing the CPU states, which will avoid calling
__hyp_get_vectors()all the time? You should be able to reuse that construct everywhere.
OK. Since I have introduced per-cpu variable, kvm_arm_hardware_enabled, against cpuidle issue, we will be able to re-use it.
Quoted 2 lines
Also, I'm not sure about
KVM_EXIT_SHUTDOWN. This looks very x86 specific (called on triple fault).
No, I don't think so.
Looking at kvm_cpu_exec() in kvm-all.c of qemu, KVM_EXIT_SHUTDOWN is handled in a generic way and results in a reset request.
On the other hand, KVM_EXIT_FAIL_ENTRY seems more arch-specific. In addition, if kvm_vcpu_ioctl() returns a negative value, run->exit_reason will never be examined.
So I think
ret -> 0
run->exit_reason -> KVM_EXIT_SHUTDOWN
or just
ret -> -ENOEXECis the best.
In either way, a guest will have no good chance to gracefully shutdown itself because we're kexec'ing (without waiting for threads' termination).
-Takahiro AKASHI
Quoted 8 lines
KVM_EXIT_FAIL_ENTRYlooks more appropriate,
and thehardware_entry_failure_reasonfield should be populated (and documented).Thanks,
M.
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 190 lines · AKASHI Takahiro
Quoted 180 lines · Marc Zyngier
Quoted 179 lines · AKASHI Takahiro
Marc,
Quoted 61 lines · Marc Zyngier
Quoted 60 lines · AKASHI Takahiro
Ashwin, Marc,
Quoted 56 lines · Marc Zyngier
Quoted 55 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 5 lines
static struct notifier_block hyp_init_cpu_pm_nb = {arch/arm/kvm/arm.c +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1124
Quoted 10 lines
arch/arm/kvm/arm.c +0
-5/* -* Execute the init code on each CPU. -*/ -on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); - -/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.Quoted 8 lines
Quoted 3 lines
Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400, I didn't notice this problem.Because GIC-400 is a GICv2 implementation, which is entirely MMIO based. GICv3 uses some system registers that are only available at EL2, and KVM needs some information contained in these registers before being able to get initialized.
I see.
Quoted 22 lines
Quoted 14 lines
Quoted 8 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?Quoted 5 lines
None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 7 lines
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.Quoted 8 lines
This seems quite complicated:
- init EL2 on all CPUs
- do some initialization
- tear all CPUs EL2 down
- let KVM drive the vectors being set or notMy questions are: why do we need to do this on *all* cpus? Can't that work on a single one?
I did initialize all the cpus partly because using preempt_enable/disable looked a bit ugly and partly because we may, in the future, do additional per-cpu initialization in
kvm_vgic_hyp_init()and/orkvm_timer_hyp_init(). But if you're comfortable with preempt_*() stuff, I don' care.Quoted 23 lines
Also, the simple fact that we were able to get some junk value is a sign that something is amiss. I'd expect a splat of some sort, because we now have a possibility of doing things in the wrong context.
Quoted 9 lines
If
kvm_hyp_call()in vgic_v3_probe()/kvm_vgic_hyp_init() is a *problem*, I hope this should work. Actually I confirmed that, with this fixup! patch, we could run a kvm guest and also successfully executed kexec on model w/gic-v3.My only concern is the following kernel message I saw when kexec shut down the kernel:
(Please note that I was running one kvm quest (pid=961) here.)Quoted 9 lines
===
sh-4.3# ./kexec -d -e
kexec version: 15.11.16.11.06-g41e52e2
arch_process_options:112:command_line: (null)
arch_process_options:114: initrd: (null)
arch_process_options:115: dtb: (null)
arch_process_options:117: port:0x0
kvm: exiting hardware virtualization
kvm [961]: Unsupported exception type: 6248304 <== this messageQuoted 4 lines
That makes me feel very uncomfortable. It looks like we've exited a guest with some horrible value in X0. How is that even possible?
This deserves to be investigated.
I guess the problem is that cpu tear-down function is called even if a kvm guest is still running in
kvm_arch_vcpu_ioctl_run().
So adding a check whether cpu has been initialized or not in every iteration ofkvm_arch_vcpu_ioctl_run()will, if necessary, terminate a guest safely without entering a guest mode. Since this check is done while interrupt is disabled, it won't interfere withkvm_arch_hardware_disable()called via IPI.
See the attached fixup patch.Again, I verified the code on model.
Thanks,
-Takahiro AKASHIQuoted 4 lines
Thanks,
M.----8<----
From77f273ba5e0c3dfcf75a5a8d1da8035cc390250cMon Sep 17 00:00:00 2001
From: AKASHI Takahiro <[email protected]>
Date: Fri, 11 Dec 2015 13:43:35 +0900
Subject: [PATCH] fixup! arm64: kvm: allows kvm cpu hotplug--- arch/arm/kvm/arm.c | 45 ++++++++++++++++++++++++++++++++++----------- 1 file changed, 34 insertions(+), 11 deletions(-)arch/arm/kvm/arm.c +2
-1@@ -573,7 +573,11 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)/* * Re-check atomic conditions */ -if (signal_pending (current)) { +if (__hyp_get_vectors () == hyp_default_vectors) { +/* cpu has been torn down */ Quoted 8 lines
arch/arm/kvm/arm.c int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run) · line 579
Quoted 2 lines
arch/arm/kvm/arm.c +2
-0+ ret = -ENOEXEC ; + run->exit_reason =KVM_EXIT_SHUTDOWN ;That feels completely overkill (and very slow). Why don't you maintain a per-cpu variable containing the CPU states, which will avoid calling
__hyp_get_vectors()all the time? You should be able to reuse that construct everywhere.
Quoted 7 lines
OK. Since I have introduced per-cpu variable,
kvm_arm_hardware_enabled, against cpuidle issue, we will be able to re-use it.Quoted 2 lines
Also, I'm not sure about
KVM_EXIT_SHUTDOWN. This looks very x86 specific (called on triple fault).No, I don't think so.
maz@approximate:~/Work/arm-platforms$ git grep KVM_EXIT_SHUTDOWN arch/x86/kvm/svm.c: kvm_run->exit_reason = KVM_EXIT_SHUTDOWN; arch/x86/kvm/vmx.c: vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; arch/x86/kvm/x86.c: vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; include/uapi/linux/kvm.h:#define KVM_EXIT_SHUTDOWN 8
And that's it. No other architecture ever generates this, and this is an undocumented API. So I'm not going to let that in until someone actually defines what this thing means.
Quoted 2 lines
Looking at
kvm_cpu_exec()inkvm-all.cof qemu,KVM_EXIT_SHUTDOWNis handled in a generic way and results in a reset request.
Which is not what we want. We want to indicate that the guest couldn't be entered. This is not due to a guest doing a triple fault (which is the way an x86 system gets rebooted).
Quoted 1 line
On the other hand,
KVM_EXIT_FAIL_ENTRYseems more arch-specific.
Certainly arch specific, but actually extremely accurate. You couldn't enter the guest, and you describe the reason in an architecture-specific fashion. This is also the only exit code that describe this exact case we're talking about here.
Quoted 5 lines
In addition, if
kvm_vcpu_ioctl()returns a negative value,run->exit_reasonwill never be examined.
So I thinkret -> 0 run->exit_reason -> KVM_EXIT_SHUTDOWN
ret = 0run->exit_reason = KVM_EXIT_FAIL_ENTRY; run->hardware_entry_failure_reason = (u64)-ENOEXEC;
Quoted 6 lines
or just
ret -> -ENOEXECis the best.
In either way, a guest will have no good chance to gracefully shutdown itself because we're kexec'ing (without waiting for threads' termination).
Well, at least userspace gets a chance - and should kexec fail, we have a chance to recover.
Thanks,
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 229 lines · Marc Zyngier
Quoted 190 lines · AKASHI Takahiro
Quoted 180 lines · Marc Zyngier
Quoted 179 lines · AKASHI Takahiro
Marc,
Quoted 61 lines · Marc Zyngier
Quoted 60 lines · AKASHI Takahiro
Ashwin, Marc,
Quoted 56 lines · Marc Zyngier
Quoted 55 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 5 lines
static struct notifier_block hyp_init_cpu_pm_nb = {(quoted diff) +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }Review static int init_hyp_mode(void) · line 1124
Quoted 10 lines
(quoted diff) +0
-5/* - * Execute the init code on each CPU. - */ - on_each_cpu(cpu_init_hyp_mode, NULL, 1); - - /* * Init HYP view of VGIC */ err = kvm_vgic_hyp_init();With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.Quoted 8 lines
Quoted 3 lines
Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400, I didn't notice this problem.Because GIC-400 is a GICv2 implementation, which is entirely MMIO based. GICv3 uses some system registers that are only available at EL2, and KVM needs some information contained in these registers before being able to get initialized.
I see.
Quoted 22 lines
Quoted 14 lines
Quoted 8 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?Quoted 5 lines
None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 7 lines
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.Quoted 8 lines
This seems quite complicated:
- init EL2 on all CPUs
- do some initialization
- tear all CPUs EL2 down
- let KVM drive the vectors being set or notMy questions are: why do we need to do this on *all* cpus? Can't that work on a single one?
I did initialize all the cpus partly because using preempt_enable/disable looked a bit ugly and partly because we may, in the future, do additional per-cpu initialization in
kvm_vgic_hyp_init()and/orkvm_timer_hyp_init(). But if you're comfortable with preempt_*() stuff, I don' care.Quoted 23 lines
Also, the simple fact that we were able to get some junk value is a sign that something is amiss. I'd expect a splat of some sort, because we now have a possibility of doing things in the wrong context.
Quoted 9 lines
If
kvm_hyp_call()in vgic_v3_probe()/kvm_vgic_hyp_init() is a *problem*, I hope this should work. Actually I confirmed that, with this fixup! patch, we could run a kvm guest and also successfully executed kexec on model w/gic-v3.My only concern is the following kernel message I saw when kexec shut down the kernel:
(Please note that I was running one kvm quest (pid=961) here.)Quoted 9 lines
===
sh-4.3# ./kexec -d -e
kexec version: 15.11.16.11.06-g41e52e2
arch_process_options:112:command_line: (null)
arch_process_options:114: initrd: (null)
arch_process_options:115: dtb: (null)
arch_process_options:117: port:0x0
kvm: exiting hardware virtualization
kvm [961]: Unsupported exception type: 6248304 <== this messageQuoted 4 lines
That makes me feel very uncomfortable. It looks like we've exited a guest with some horrible value in X0. How is that even possible?
This deserves to be investigated.
I guess the problem is that cpu tear-down function is called even if a kvm guest is still running in
kvm_arch_vcpu_ioctl_run().
So adding a check whether cpu has been initialized or not in every iteration ofkvm_arch_vcpu_ioctl_run()will, if necessary, terminate a guest safely without entering a guest mode. Since this check is done while interrupt is disabled, it won't interfere withkvm_arch_hardware_disable()called via IPI.
See the attached fixup patch.Again, I verified the code on model.
Thanks,
-Takahiro AKASHIQuoted 4 lines
Thanks,
M.----8<----
From77f273ba5e0c3dfcf75a5a8d1da8035cc390250cMon Sep 17 00:00:00 2001
From: AKASHI Takahiro <[email protected]>
Date: Fri, 11 Dec 2015 13:43:35 +0900
Subject: [PATCH] fixup! arm64: kvm: allows kvm cpu hotplug--- arch/arm/kvm/arm.c | 45 ++++++++++++++++++++++++++++++++++----------- 1 file changed, 34 insertions(+), 11 deletions(-)arch/arm/kvm/arm.c +2
-1@@ -573,7 +573,11 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)/* * Re-check atomic conditions */ -if (signal_pending (current)) { +if (__hyp_get_vectors () == hyp_default_vectors) { +/* cpu has been torn down */ Quoted 8 lines
arch/arm/kvm/arm.c int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run) · line 579
Quoted 2 lines
arch/arm/kvm/arm.c +2
-0+ ret = -ENOEXEC ; + run->exit_reason =KVM_EXIT_SHUTDOWN ;That feels completely overkill (and very slow). Why don't you maintain a per-cpu variable containing the CPU states, which will avoid calling
__hyp_get_vectors()all the time? You should be able to reuse that construct everywhere.Quoted 7 lines
OK. Since I have introduced per-cpu variable,
kvm_arm_hardware_enabled, against cpuidle issue, we will be able to re-use it.Quoted 2 lines
Also, I'm not sure about
KVM_EXIT_SHUTDOWN. This looks very x86 specific (called on triple fault).No, I don't think so.
maz at approximate:~/Work/arm-platforms$ git grep
KVM_EXIT_SHUTDOWNarch/x86/kvm/svm.c: kvm_run->exit_reason = KVM_EXIT_SHUTDOWN; arch/x86/kvm/vmx.c: vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; arch/x86/kvm/x86.c: vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN;include/uapi/linux/kvm.h:#define
KVM_EXIT_SHUTDOWN8And that's it. No other architecture ever generates this, and this is an undocumented API. So I'm not going to let that in until someone actually defines what this thing means.
Quoted 2 lines
Looking at
kvm_cpu_exec()inkvm-all.cof qemu,KVM_EXIT_SHUTDOWNis handled in a generic way and results in a reset request.Which is not what we want. We want to indicate that the guest couldn't be entered. This is not due to a guest doing a triple fault (which is the way an x86 system gets rebooted).
Quoted 1 line
On the other hand,
KVM_EXIT_FAIL_ENTRYseems more arch-specific.Certainly arch specific, but actually extremely accurate. You couldn't enter the guest, and you describe the reason in an architecture-specific fashion. This is also the only exit code that describe this exact case we're talking about here.
Quoted 5 lines
In addition, if
kvm_vcpu_ioctl()returns a negative value,run->exit_reasonwill never be examined.
So I thinkret -> 0 run->exit_reason -> KVM_EXIT_SHUTDOWN
Quoted 3 lines
ret = 0
run->exit_reason=KVM_EXIT_FAIL_ENTRY;run->hardware_entry_failure_reason= (u64)-ENOEXEC;
OK.
Quoted 9 lines
Quoted 6 lines
or just
ret -> -ENOEXECis the best.
In either way, a guest will have no good chance to gracefully shutdown itself because we're kexec'ing (without waiting for threads' termination).
Well, at least userspace gets a chance - and should kexec fail, we have a chance to recover.
Well, the current kexec implementation (on arm64) never fails except very early stage :)
So please review the attached fixup patch, again.
Thanks,
-Takahiro AKASHI
Quoted 4 lines
Thanks,
M.
----8<----
From ec6c07fe80d6ba96855468f61daffa9b91cf5622 Mon Sep 17 00:00:00 2001
From: AKASHI Takahiro <[email protected]>
Date: Fri, 11 Dec 2015 13:43:35 +0900
Subject: [PATCH] fixup! arm64: kvm: allows kvm cpu hotplug
--- arch/arm/kvm/arm.c | 62 +++++++++++++++++++++++++++++++++++----------------- 1 file changed, 42 insertions(+), 20 deletions(-)
arch/arm/kvm/arm.c +42 -20
@@ -573,7 +573,13 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)/* * Re-check atomic conditions */ -if (signal_pending (current)) { +if (!__this_cpu_read (kvm_arm_hardware_enabled)) { +/* cpu has been torn down */ + ret =0 ; + run->exit_reason =KVM_EXIT_FAIL_ENTRY ; + run->fail_entry.hardware_entry_failure_reason + = (u64)-ENOEXEC ; + }else if (signal_pending (current)) { ret = -EINTR ; run->exit_reason =KVM_EXIT_INTR ; }
@@ -950,7 +956,7 @@ long kvm_arch_vm_ioctl(struct file *filp, } } -int kvm_arch_hardware_enable (void ) +static void cpu_init_hyp_mode (void ) {phys_addr_t boot_pgd_ptr;phys_addr_t pgd_ptr;
@@ -958,9 +964,6 @@ int kvm_arch_hardware_enable(void)unsigned long stack_page;unsigned long vector_ptr; -if (__hyp_get_vectors () != hyp_default_vectors) -return 0 ; -/* Switch from the HYP stub to our own HYP init vector */ __hyp_set_vectors (kvm_get_idmap_vector ());
@@ -973,24 +976,38 @@ int kvm_arch_hardware_enable(void)__cpu_init_hyp_mode (boot_pgd_ptr, pgd_ptr, hyp_stack_ptr, vector_ptr);kvm_arm_init_debug (); - -return 0 ; } -void kvm_arch_hardware_disable (void ) +static void cpu_reset_hyp_mode (void ) {phys_addr_t boot_pgd_ptr;phys_addr_t phys_idmap_start; -if (__hyp_get_vectors () == hyp_default_vectors) -return ; - boot_pgd_ptr =kvm_mmu_get_boot_httbr (); phys_idmap_start =kvm_get_idmap_start ();__cpu_reset_hyp_mode (boot_pgd_ptr, phys_idmap_start); } +int kvm_arch_hardware_enable (void ) +{ +if (!__this_cpu_read (kvm_arm_hardware_enabled)) { +cpu_init_hyp_mode (); +__this_cpu_write (kvm_arm_hardware_enabled,1 ); + } + +return 0 ; +} + +void kvm_arch_hardware_disable (void ) +{ +if (!__this_cpu_read (kvm_arm_hardware_enabled)) +return ; + +cpu_reset_hyp_mode (); +__this_cpu_write (kvm_arm_hardware_enabled,0 ); +} +#ifdef CONFIG_CPU_PM static int hyp_init_cpu_pm_notifier(struct notifier_block *self, unsigned long cmd,
@@ -998,19 +1015,13 @@ static int hyp_init_cpu_pm_notifier(struct notifier_block *self, {switch (cmd) {case CPU_PM_ENTER : -if (__hyp_get_vectors () != hyp_default_vectors) -__this_cpu_write (kvm_arm_hardware_enabled,1 ); -else -__this_cpu_write (kvm_arm_hardware_enabled,0 ); -/* -* don't call kvm_arch_hardware_disable() in case of -* CPU_PM_ENTER because it does't actually save any state. -*/ +if (__this_cpu_read (kvm_arm_hardware_enabled)) +cpu_reset_hyp_mode ();return NOTIFY_OK ;case CPU_PM_EXIT :if (__this_cpu_read (kvm_arm_hardware_enabled)) -kvm_arch_hardware_enable (); +cpu_init_hyp_mode ();return NOTIFY_OK ;
@@ -1114,9 +1125,20 @@ static int init_hyp_mode(void) }/* +* Init this CPU temporarily to execute kvm_hyp_call() +* during kvm_vgic_hyp_init(). +*/ +preempt_disable (); +cpu_init_hyp_mode (); + +/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init (); + +cpu_reset_hyp_mode (); +preempt_enable (); +if (err) goto out_free_context;
Signature · 1 line
1.7.9.5
Re: [PATCH v12 04/16] arm64: kvm: allows kvm cpu hotplug
Quoted 275 lines · AKASHI Takahiro
Quoted 229 lines · Marc Zyngier
Quoted 190 lines · AKASHI Takahiro
Quoted 180 lines · Marc Zyngier
Quoted 179 lines · AKASHI Takahiro
Marc,
Quoted 61 lines · Marc Zyngier
Quoted 60 lines · AKASHI Takahiro
Ashwin, Marc,
Quoted 56 lines · Marc Zyngier
Quoted 55 lines · Ashwin Chaugule
Hello,
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
The current kvm implementation on arm64 does cpu-specific initialization at system boot, and has no way to gracefully shutdown a core in terms of kvm. This prevents, especially, kexec from rebooting the system on a boot core in EL2.
This patch adds a cpu tear-down function and also puts an existing cpu-init code into a separate function,
kvm_arch_hardware_disable()andkvm_arch_hardware_enable()respectively.
We don't need arm64-specific cpu hotplug hook any more.Since this patch modifies common part of code between arm and arm64, one stub definition,
__cpu_reset_hyp_mode(), is added on arm side to avoid compiling errors.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm/include/asm/kvm_host.h | 10 ++++- arch/arm/include/asm/kvm_mmu.h | 1 + arch/arm/kvm/arm.c | 79 ++++++++++++++++++--------------------- arch/arm/kvm/mmu.c | 5 +++ arch/arm64/include/asm/kvm_host.h | 16 +++++++- arch/arm64/include/asm/kvm_mmu.h | 1 + arch/arm64/include/asm/virt.h | 9 +++++ arch/arm64/kvm/hyp-init.S | 33 ++++++++++++++++ arch/arm64/kvm/hyp.S | 32 ++++++++++++++-- 9 files changed, 138 insertions(+), 48 deletions(-)[..]
Quoted 5 lines
static struct notifier_block hyp_init_cpu_pm_nb = {arch/arm/kvm/arm.c +0
-0@@ -1108,11 +1119,6 @@ static int init_hyp_mode(void) }arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1124
Quoted 10 lines
arch/arm/kvm/arm.c +0
-5/* -* Execute the init code on each CPU. -*/ -on_each_cpu (cpu_init_hyp_mode,NULL ,1 ); - -/* * Init HYP view of VGIC */ err =kvm_vgic_hyp_init ();With this flow, the
cpu_init_hyp_mode()is called only at VM guest creation, butvgic_hyp_init()is called at bootup. On a system with GICv3, it looks like we end up with bogus values from theICH_VTR_EL2(to get the number of LRs), because we're not reading it from EL2 anymore.Quoted 8 lines
Quoted 3 lines
Thank you for pointing this out.
Recently I tested my kdump code on hikey, and as hikey(hi6220) has gic-400, I didn't notice this problem.Because GIC-400 is a GICv2 implementation, which is entirely MMIO based. GICv3 uses some system registers that are only available at EL2, and KVM needs some information contained in these registers before being able to get initialized.
I see.
Quoted 22 lines
Quoted 14 lines
Quoted 8 lines
Indeed, this is completely broken (I just reproduced the issue on a model). I wish this kind of details had been checked earlier, but thanks for pointing it out.
Quoted 3 lines
Whats the best way to fix this?
- Callkvm_arch_hardware_enable()beforevgic_hyp_init()and disable later?
- Fold the VGIC init stuff back intohardware_enable()?Quoted 5 lines
None of that works -
kvm_arch_hardware_enable()is called once per CPU, whilevgic_hyp_init()can only be called once. Also,kvm_arch_hardware_enable()is called from interrupt context, and I wouldn't feel comfortable starting probing DT and allocating stuff from there.Quoted 7 lines
Do you think so?
How about the fixup! patch attached below?
The point is that, like Ashwin's first idea, we initialize cpus temporarily beforekvm_vgic_hyp_init()and then soon reset cpus again. Thus, kvm cpu hotplug will still continue to work as before.
Now thatcpu_init_hyp_mode()is revived as exactly the same as Marc's original code, the change will not be a big jump.Quoted 8 lines
This seems quite complicated:
- init EL2 on all CPUs
- do some initialization
- tear all CPUs EL2 down
- let KVM drive the vectors being set or notMy questions are: why do we need to do this on *all* cpus? Can't that work on a single one?
I did initialize all the cpus partly because using preempt_enable/disable looked a bit ugly and partly because we may, in the future, do additional per-cpu initialization in
kvm_vgic_hyp_init()and/orkvm_timer_hyp_init(). But if you're comfortable with preempt_*() stuff, I don' care.Quoted 23 lines
Also, the simple fact that we were able to get some junk value is a sign that something is amiss. I'd expect a splat of some sort, because we now have a possibility of doing things in the wrong context.
Quoted 9 lines
If
kvm_hyp_call()in vgic_v3_probe()/kvm_vgic_hyp_init() is a *problem*, I hope this should work. Actually I confirmed that, with this fixup! patch, we could run a kvm guest and also successfully executed kexec on model w/gic-v3.My only concern is the following kernel message I saw when kexec shut down the kernel:
(Please note that I was running one kvm quest (pid=961) here.)Quoted 9 lines
===
sh-4.3# ./kexec -d -e
kexec version: 15.11.16.11.06-g41e52e2
arch_process_options:112:command_line: (null)
arch_process_options:114: initrd: (null)
arch_process_options:115: dtb: (null)
arch_process_options:117: port:0x0
kvm: exiting hardware virtualization
kvm [961]: Unsupported exception type: 6248304 <== this messageQuoted 4 lines
That makes me feel very uncomfortable. It looks like we've exited a guest with some horrible value in X0. How is that even possible?
This deserves to be investigated.
I guess the problem is that cpu tear-down function is called even if a kvm guest is still running in
kvm_arch_vcpu_ioctl_run().
So adding a check whether cpu has been initialized or not in every iteration ofkvm_arch_vcpu_ioctl_run()will, if necessary, terminate a guest safely without entering a guest mode. Since this check is done while interrupt is disabled, it won't interfere withkvm_arch_hardware_disable()called via IPI.
See the attached fixup patch.Again, I verified the code on model.
Thanks,
-Takahiro AKASHIQuoted 4 lines
Thanks,
M.----8<----
From77f273ba5e0c3dfcf75a5a8d1da8035cc390250cMon Sep 17 00:00:00 2001
From: AKASHI Takahiro <[email protected]>
Date: Fri, 11 Dec 2015 13:43:35 +0900
Subject: [PATCH] fixup! arm64: kvm: allows kvm cpu hotplug--- arch/arm/kvm/arm.c | 45 ++++++++++++++++++++++++++++++++++----------- 1 file changed, 34 insertions(+), 11 deletions(-)arch/arm/kvm/arm.c +2
-1@@ -573,7 +573,11 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)/* * Re-check atomic conditions */ -if (signal_pending (current)) { +if (__hyp_get_vectors () == hyp_default_vectors) { +/* cpu has been torn down */ Quoted 8 lines
arch/arm/kvm/arm.c int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run) · line 579
Quoted 2 lines
arch/arm/kvm/arm.c +2
-0+ ret = -ENOEXEC ; + run->exit_reason =KVM_EXIT_SHUTDOWN ;That feels completely overkill (and very slow). Why don't you maintain a per-cpu variable containing the CPU states, which will avoid calling
__hyp_get_vectors()all the time? You should be able to reuse that construct everywhere.Quoted 7 lines
OK. Since I have introduced per-cpu variable,
kvm_arm_hardware_enabled, against cpuidle issue, we will be able to re-use it.Quoted 2 lines
Also, I'm not sure about
KVM_EXIT_SHUTDOWN. This looks very x86 specific (called on triple fault).No, I don't think so.
maz@approximate:~/Work/arm-platforms$ git grep KVM_EXIT_SHUTDOWN arch/x86/kvm/svm.c: kvm_run->exit_reason = KVM_EXIT_SHUTDOWN; arch/x86/kvm/vmx.c: vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; arch/x86/kvm/x86.c: vcpu->run->exit_reason = KVM_EXIT_SHUTDOWN; include/uapi/linux/kvm.h:#define KVM_EXIT_SHUTDOWN 8And that's it. No other architecture ever generates this, and this is an undocumented API. So I'm not going to let that in until someone actually defines what this thing means.
Quoted 2 lines
Looking at
kvm_cpu_exec()inkvm-all.cof qemu,KVM_EXIT_SHUTDOWNis handled in a generic way and results in a reset request.Which is not what we want. We want to indicate that the guest couldn't be entered. This is not due to a guest doing a triple fault (which is the way an x86 system gets rebooted).
Quoted 1 line
On the other hand,
KVM_EXIT_FAIL_ENTRYseems more arch-specific.Certainly arch specific, but actually extremely accurate. You couldn't enter the guest, and you describe the reason in an architecture-specific fashion. This is also the only exit code that describe this exact case we're talking about here.
Quoted 5 lines
In addition, if
kvm_vcpu_ioctl()returns a negative value,run->exit_reasonwill never be examined.
So I thinkret -> 0 run->exit_reason -> KVM_EXIT_SHUTDOWNQuoted 3 lines
ret = 0
run->exit_reason=KVM_EXIT_FAIL_ENTRY;run->hardware_entry_failure_reason= (u64)-ENOEXEC;OK.
Quoted 9 lines
Quoted 6 lines
or just
ret -> -ENOEXECis the best.
In either way, a guest will have no good chance to gracefully shutdown itself because we're kexec'ing (without waiting for threads' termination).
Well, at least userspace gets a chance - and should kexec fail, we have a chance to recover.
Well, the current kexec implementation (on arm64) never fails except very early stage :)
So please review the attached fixup patch, again.
Thanks,
-Takahiro AKASHIQuoted 4 lines
Thanks,
M.----8<----
Fromec6c07fe80d6ba96855468f61daffa9b91cf5622Mon Sep 17 00:00:00 2001
From: AKASHI Takahiro <[email protected]>
Date: Fri, 11 Dec 2015 13:43:35 +0900
Subject: [PATCH] fixup! arm64: kvm: allows kvm cpu hotplug--- arch/arm/kvm/arm.c | 62 +++++++++++++++++++++++++++++++++++----------------- 1 file changed, 42 insertions(+), 20 deletions(-)arch/arm/kvm/arm.c +0
-0
arch/arm/kvm/arm.c int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run) · line 576
Quoted 5 lines
arch/arm/kvm/arm.c +1
-1@@ -573,7 +573,13 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)/* * Re-check atomic conditions */ -if (signal_pending (current)) { +if (!__this_cpu_read (kvm_arm_hardware_enabled)) {
if (unlikely(...))
Quoted 59 lines
arch/arm/kvm/arm.c +16
-10+/* cpu has been torn down */ + ret =0 ; + run->exit_reason =KVM_EXIT_FAIL_ENTRY ; + run->fail_entry.hardware_entry_failure_reason + = (u64)-ENOEXEC ; + }else if (signal_pending (current)) { ret = -EINTR ; run->exit_reason =KVM_EXIT_INTR ; }@@ -950,7 +956,7 @@ long kvm_arch_vm_ioctl(struct file *filp, } } -int kvm_arch_hardware_enable (void ) +static void cpu_init_hyp_mode (void ) {phys_addr_t boot_pgd_ptr;phys_addr_t pgd_ptr;@@ -958,9 +964,6 @@ int kvm_arch_hardware_enable(void)unsigned long stack_page;unsigned long vector_ptr; -if (__hyp_get_vectors () != hyp_default_vectors) -return 0 ; -/* Switch from the HYP stub to our own HYP init vector */ __hyp_set_vectors (kvm_get_idmap_vector ());@@ -973,24 +976,38 @@ int kvm_arch_hardware_enable(void)__cpu_init_hyp_mode (boot_pgd_ptr, pgd_ptr, hyp_stack_ptr, vector_ptr);kvm_arm_init_debug (); - -return 0 ; } -void kvm_arch_hardware_disable (void ) +static void cpu_reset_hyp_mode (void ) {phys_addr_t boot_pgd_ptr;phys_addr_t phys_idmap_start; -if (__hyp_get_vectors () == hyp_default_vectors) -return ; - boot_pgd_ptr =kvm_mmu_get_boot_httbr (); phys_idmap_start =kvm_get_idmap_start ();__cpu_reset_hyp_mode (boot_pgd_ptr, phys_idmap_start); } +int kvm_arch_hardware_enable (void ) +{ +if (!__this_cpu_read (kvm_arm_hardware_enabled)) { +cpu_init_hyp_mode (); +__this_cpu_write (kvm_arm_hardware_enabled,1 ); + } + +return 0 ;
arch/arm/kvm/arm.c int kvm_arch_hardware_enable(void) · line 1009
Quoted 10 lines
arch/arm/kvm/arm.c +10
-0+} + +void kvm_arch_hardware_disable (void ) +{ +if (!__this_cpu_read (kvm_arm_hardware_enabled)) +return ; + +cpu_reset_hyp_mode (); +__this_cpu_write (kvm_arm_hardware_enabled,0 ); +}
Are you guaranteed that these two functions are called from a context where both preemption and interrupts are disabled? If not, you cannot use the __this_cpu* operations, and you must stick with this_cpu*.
Quoted 14 lines
arch/arm/kvm/arm.c +1
-6+#ifdef CONFIG_CPU_PM static int hyp_init_cpu_pm_notifier(struct notifier_block *self, unsigned long cmd,@@ -998,19 +1015,13 @@ static int hyp_init_cpu_pm_notifier(struct notifier_block *self, {switch (cmd) {case CPU_PM_ENTER : -if (__hyp_get_vectors () != hyp_default_vectors) -__this_cpu_write (kvm_arm_hardware_enabled,1 ); -else -__this_cpu_write (kvm_arm_hardware_enabled,0 ); -/* -* don't call kvm_arch_hardware_disable() in case of
arch/arm/kvm/arm.c static int hyp_init_cpu_pm_notifier(struct notifier_block *self, · line 1024
Quoted 10 lines
arch/arm/kvm/arm.c +3
-3-* CPU_PM_ENTER because it does't actually save any state. -*/ +if (__this_cpu_read (kvm_arm_hardware_enabled)) +cpu_reset_hyp_mode ();return NOTIFY_OK ;case CPU_PM_EXIT :if (__this_cpu_read (kvm_arm_hardware_enabled)) -kvm_arch_hardware_enable (); +cpu_init_hyp_mode ();
It would be good to have a comment as to why we're not directly calling hardware_enable/disable.
Quoted 14 lines
arch/arm/kvm/arm.c +7
-0return NOTIFY_OK ;@@ -1114,9 +1125,20 @@ static int init_hyp_mode(void) }/* +* Init this CPU temporarily to execute kvm_hyp_call() +* during kvm_vgic_hyp_init(). +*/ +preempt_disable (); +cpu_init_hyp_mode (); + +/*
arch/arm/kvm/arm.c static int init_hyp_mode(void) · line 1144
Quoted 10 lines
arch/arm/kvm/arm.c +4
-0* Init HYP view of VGIC */ err =kvm_vgic_hyp_init (); + +cpu_reset_hyp_mode (); +preempt_enable (); +if (err) goto out_free_context;
Thanks,
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 5/16 arm64: Add back cpu_reset routines
In reply to Geoff LevandCommit 68234df4ea7939f98431aa81113fbdce10c4a84b (arm64: kill flush_cache_all()) removed the global arm64 routines cpu_reset() and cpu_soft_restart() needed by the arm64 kexec and kdump support. Add simplified versions of those two routines back with some changes needed for kexec in the new files cpu_reset.S, and cpu_reset.h.
When a CPU is reset it needs to be put into the exception level it had when it entered the kernel. Update cpu_reset() to accept an argument which signals if the reset address needs to be entered at EL1 or EL2.
--- arch/arm64/kernel/cpu-reset.S | 84 +++++++++++++++++++++++++++++++++++++++++++ arch/arm64/kernel/cpu-reset.h | 22 ++++++++++++ 2 files changed, 106 insertions(+) create mode 100644 arch/arm64/kernel/cpu-reset.S create mode 100644 arch/arm64/kernel/cpu-reset.h
arch/arm64/kernel/cpu-reset.S new file +84 -0
@@ -0,0 +1,84 @@ +/* + * cpu reset routines + * + * Copyright (C) 2001 Deep Blue Solutions Ltd. + * Copyright (C) 2012 ARM Ltd. + * Copyright (C) 2015 Huawei Futurewei Technologies. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ + +#include <linux/linkage.h> +#include <asm/assembler.h> +#include <asm/virt.h> + +#define SCTLR_ELx_I (1 << 12) +#define SCTLR_ELx_SA (1 << 3) +#define SCTLR_ELx_C (1 << 2) +#define SCTLR_ELx_A (1 << 1) +#define SCTLR_ELx_M 1 +#define SCTLR_ELx_FLAGS (SCTLR_ELx_M | SCTLR_ELx_A | SCTLR_ELx_C | \ + SCTLR_ELx_SA | SCTLR_ELx_I) + +.text +.pushsection .idmap.text, "ax" + +.align 5 + +/* + * cpu_soft_restart(cpu_reset, el2_switch, entry, arg0, arg1) - Perform a cpu + * soft reset. + * + * @cpu_reset: Physical address of the cpu_reset routine. + * @el2_switch: Flag to indicate a swich to EL2 is needed, passed to cpu_reset. + * @entry: Location to jump to for soft reset, passed to cpu_reset. + * arg0: First argument passed to @entry. + * arg1: Second argument passed to @entry. + */ + +ENTRY(cpu_soft_restart) + mov x18, x0 // cpu_reset + mov x0, x1 // el2_switch + mov x1, x2 // entry + mov x2, x3 // arg0 + mov x3, x4 // arg1 + ret x18 +ENDPROC(cpu_soft_restart) + +/* + * cpu_reset(el2_switch, entry, arg0, arg1) - Helper for cpu_soft_restart. + * + * @el2_switch: Flag to indicate a swich to EL2 is needed. + * @entry: Location to jump to for soft reset. + * arg0: First argument passed to @entry. + * arg1: Second argument passed to @entry. + * + * Put the CPU into the same state as it would be if it had been reset, and + * branch to what would be the reset vector. It must be executed with the + * flat identity mapping. + */ + +ENTRY(cpu_reset) + /* Clear sctlr_el1 flags. */ + mrs x12, sctlr_el1 + ldr x13, =SCTLR_ELx_FLAGS + bic x12, x12, x13 + msr sctlr_el1, x12 + isb + + cbz x0, 1f // el2_switch? + + mov x0, x1 // entry + mov x1, x2 // arg0 + mov x2, x3 // arg1 + hvc #HVC_CALL_FUNC // no return + +1: mov x18, x1 // entry + mov x0, x2 // arg0 + mov x1, x3 // arg1 + ret x18 +ENDPROC(cpu_reset) + +.popsection
arch/arm64/kernel/cpu-reset.h new file +22 -0
@@ -0,0 +1,22 @@ +/* +* cpu reset routines +* +* Copyright (C) 2015 Huawei Futurewei Technologies. +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License version 2 as +* published by the Free Software Foundation. +*/ + +#if !defined(_ARM64_CPU_RESET_H) +#define _ARM64_CPU_RESET_H + +#include <asm/virt.h> + +void __attribute__((noreturn))cpu_reset (unsigned long el2_switch, +unsigned long entry,unsigned long arg0,unsigned long arg1); +void __attribute__((noreturn))cpu_soft_restart (phys_addr_t cpu_reset, +unsigned long el2_switch,unsigned long entry,unsigned long arg0, +unsigned long arg1); + +#endif
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 05/16] arm64: Add back cpu_reset routines
Quoted 35 lines · Geoff Levand
Commit
68234df4ea7939f98431aa81113fbdce10c4a84b(arm64: killflush_cache_all()) removed the global arm64 routinescpu_reset()andcpu_soft_restart()needed by the arm64 kexec and kdump support. Add simplified versions of those two routines back with some changes needed for kexec in the new filescpu_reset.S, andcpu_reset.h.When a CPU is reset it needs to be put into the exception level it had when it entered the kernel. Update
cpu_reset()to accept an argument which signals if the reset address needs to be entered at EL1 or EL2.Signed-off-by Geoff Levand <[email protected]>--- arch/arm64/kernel/cpu-reset.S | 84 +++++++++++++++++++++++++++++++++++++++++++ arch/arm64/kernel/cpu-reset.h | 22 ++++++++++++ 2 files changed, 106 insertions(+) create mode 100644 arch/arm64/kernel/cpu-reset.S create mode 100644 arch/arm64/kernel/cpu-reset.harch/arm64/kernel/cpu-reset.S new file +11
-0@@ -0,0 +1,84 @@ +/* + * cpu reset routines + * + * Copyright (C) 2001 Deep Blue Solutions Ltd. + * Copyright (C) 2012 ARM Ltd. + * Copyright (C) 2015 Huawei Futurewei Technologies. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */
arch/arm64/kernel/cpu-reset.S line 21
Quoted 10 lines
arch/arm64/kernel/cpu-reset.S new file +10
-0+ +#include <linux/linkage.h> +#include <asm/assembler.h> +#include <asm/virt.h> + +#define SCTLR_ELx_I (1 << 12) +#define SCTLR_ELx_SA (1 << 3) +#define SCTLR_ELx_C (1 << 2) +#define SCTLR_ELx_A (1 << 1) +#define SCTLR_ELx_M 1
kvm_arm.h already has some of these defined, and sysreg.h has another bunch. Can you please consolidate this into a single place (probably sysreg.h)?
arch/arm64/kernel/cpu-reset.S line 28
Quoted 7 lines
arch/arm64/kernel/cpu-reset.S new file +7
-0+#define SCTLR_ELx_FLAGS (SCTLR_ELx_M | SCTLR_ELx_A | SCTLR_ELx_C | \ + SCTLR_ELx_SA | SCTLR_ELx_I) + +.text +.pushsection .idmap.text, "ax" + +.align 5
Why is this .align directive required?
Quoted 10 lines
arch/arm64/kernel/cpu-reset.S new file +10
-0+ +/* + * cpu_soft_restart(cpu_reset, el2_switch, entry, arg0, arg1) - Perform a cpu + * soft reset. + * + * @cpu_reset: Physical address of the cpu_reset routine. + * @el2_switch: Flag to indicate a swich to EL2 is needed, passed to cpu_reset. + * @entry: Location to jump to for soft reset, passed to cpu_reset. + * arg0: First argument passed to @entry. + * arg1: Second argument passed to @entry.
arch/arm64/kernel/cpu-reset.S line 48
Quoted 10 lines
arch/arm64/kernel/cpu-reset.S new file +10
-0+ */ + +ENTRY(cpu_soft_restart) + mov x18, x0 // cpu_reset + mov x0, x1 // el2_switch + mov x1, x2 // entry + mov x2, x3 // arg0 + mov x3, x4 // arg1 + ret x18 +ENDPROC(cpu_soft_restart)
Grepping through the tree, I can only find a single use of cpu_soft_restart, with cpu_reset as its first parameter.
Why do we need this indirection? Having
void cpu_soft_restart(el2_switch, entry, arg0, arg1);
should be enough...
Quoted 24 lines
arch/arm64/kernel/cpu-reset.S new file +24
-0+ +/* + * cpu_reset(el2_switch, entry, arg0, arg1) - Helper for cpu_soft_restart. + * + * @el2_switch: Flag to indicate a swich to EL2 is needed. + * @entry: Location to jump to for soft reset. + * arg0: First argument passed to @entry. + * arg1: Second argument passed to @entry. + * + * Put the CPU into the same state as it would be if it had been reset, and + * branch to what would be the reset vector. It must be executed with the + * flat identity mapping. + */ + +ENTRY(cpu_reset) + /* Clear sctlr_el1 flags. */ + mrs x12, sctlr_el1 + ldr x13, =SCTLR_ELx_FLAGS + bic x12, x12, x13 + msr sctlr_el1, x12 + isb + + cbz x0, 1f // el2_switch? +
arch/arm64/kernel/cpu-reset.S line 82
Quoted 10 lines
arch/arm64/kernel/cpu-reset.S new file +10
-0+ mov x0, x1 // entry + mov x1, x2 // arg0 + mov x2, x3 // arg1 + hvc #HVC_CALL_FUNC // no return + +1: mov x18, x1 // entry + mov x0, x2 // arg0 + mov x1, x3 // arg1 + ret x18 +ENDPROC(cpu_reset)
And this can be an actual part of cpu_soft_reset.
Quoted 9 lines
arch/arm64/kernel/cpu-reset.S new file +2
-0+ +.popsectionarch/arm64/kernel/cpu-reset.h new file +1
-0@@ -0,0 +1,22 @@ +/*
arch/arm64/kernel/cpu-reset.h line 11
Quoted 10 lines
arch/arm64/kernel/cpu-reset.h new file +10
-0+* cpu reset routines +* +* Copyright (C) 2015 Huawei Futurewei Technologies. +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License version 2 as +* published by the Free Software Foundation. +*/ + +#if !defined(_ARM64_CPU_RESET_H)
The usual idiom is "#ifndef __BLAH_H".
arch/arm64/kernel/cpu-reset.h line 22
Quoted 12 lines
arch/arm64/kernel/cpu-reset.h new file +11
-0+#define _ARM64_CPU_RESET_H + +#include <asm/virt.h> + +void __attribute__((noreturn))cpu_reset (unsigned long el2_switch, +unsigned long entry,unsigned long arg0,unsigned long arg1); +void __attribute__((noreturn))cpu_soft_restart (phys_addr_t cpu_reset, +unsigned long el2_switch,unsigned long entry,unsigned long arg0, +unsigned long arg1); + +#endif
There is a __noreturn attribute already defined.
Thanks,
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 05/16] arm64: Add back cpu_reset routines
Quoted 13 lines · Marc Zyngier
arch/arm64/kernel/cpu-reset.S line 48
Quoted 8 lines · Geoff Levand
arch/arm64/kernel/cpu-reset.S new file +8
-0+ENTRY(cpu_soft_restart) + mov x18, x0 // cpu_reset + mov x0, x1 // el2_switch + mov x1, x2 // entry + mov x2, x3 // arg0 + mov x3, x4 // arg1 + ret x18 +ENDPROC(cpu_soft_restart)Grepping through the tree, I can only find a single use of
cpu_soft_restart, withcpu_resetas its first parameter.
Quoted 1 line
Why do we need this indirection? Having
It is needed because we need to execute cpu_reset() in physical address space.
Quoted 4 lines
void
cpu_soft_restart(el2_switch, entry, arg0, arg1);should be enough...
We can do with only cpu_soft_restart(), but then a function pointer to __pa() of it need to be called. May be current approach is more cleaner.
~Pratyush
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 05/16] arm64: Add back cpu_reset routines
Quoted 19 lines · On Mon, 30 Nov 2015 10:58:43 +0530 Pratyush Anand <[email protected]> wrote:
Quoted 13 lines · Marc Zyngier
Review
Quoted 8 lines · Geoff Levand
(quoted diff) +8
-0+ENTRY(cpu_soft_restart) + mov x18, x0 // cpu_reset + mov x0, x1 // el2_switch + mov x1, x2 // entry + mov x2, x3 // arg0 + mov x3, x4 // arg1 + ret x18 +ENDPROC(cpu_soft_restart)Grepping through the tree, I can only find a single use of
cpu_soft_restart, withcpu_resetas its first parameter.Quoted 1 line
Why do we need this indirection? Having
It is needed because we need to execute
cpu_reset()in physical address space.Quoted 1 line
Quoted 6 lines
Quoted 3 lines
void
cpu_soft_restart(el2_switch, entry, arg0, arg1);should be enough...
We can do with only
cpu_soft_restart(), but then a function pointer to__pa()of it need to be called. May be current approach is more cleaner.
All that can be solved in C, and mostly at compile time. Using an assembler trampoline is complicating things for no good reason:
arch/arm64/kernel/cpu-reset.S +3 -23
@@ -28,27 +28,7 @@ .align 5 /* - * cpu_soft_restart(cpu_reset, el2_switch, entry, arg0, arg1) - Perform a cpu - * soft reset. - * - * @cpu_reset: Physical address of the cpu_reset routine. - * @el2_switch: Flag to indicate a swich to EL2 is needed, passed to cpu_reset. - * @entry: Location to jump to for soft reset, passed to cpu_reset. - * arg0: First argument passed to @entry. - * arg1: Second argument passed to @entry. - */ - -ENTRY(cpu_soft_restart) - mov x18, x0 // cpu_reset - mov x0, x1 // el2_switch - mov x1, x2 // entry - mov x2, x3 // arg0 - mov x3, x4 // arg1 - ret x18 -ENDPROC(cpu_soft_restart) - -/* - * cpu_reset(el2_switch, entry, arg0, arg1) - Helper for cpu_soft_restart. + * __cpu_soft_restart(el2_switch, entry, arg0, arg1) - Helper for cpu_soft_restart. * * @el2_switch: Flag to indicate a swich to EL2 is needed. * @entry: Location to jump to for soft reset.
@@ -60,7 +40,7 @@ ENDPROC(cpu_soft_restart) * flat identity mapping. */ -ENTRY(cpu_reset) +ENTRY(__cpu_soft_restart) /* Clear sctlr_el1 flags. */ mrs x12, sctlr_el1 ldr x13, =SCTLR_ELx_FLAGS
@@ -79,6 +59,6 @@ ENTRY(cpu_reset) mov x0, x2 // arg0 mov x1, x3 // arg1 ret x18 -ENDPROC(cpu_reset) +ENDPROC(__cpu_soft_restart) .popsection
arch/arm64/kernel/cpu-reset.h +12 -5
@@ -13,10 +13,17 @@#include <asm/virt.h> -void __attribute__((noreturn))cpu_reset (unsigned long el2_switch, -unsigned long entry,unsigned long arg0,unsigned long arg1); -void __attribute__((noreturn))cpu_soft_restart (phys_addr_t cpu_reset, -unsigned long el2_switch,unsigned long entry,unsigned long arg0, -unsigned long arg1); +void __cpu_soft_restart (unsigned long el2_switch,unsigned long entry, +unsigned long arg0,unsigned long arg1); + +static inline +void __noreturn cpu_soft_restart (unsigned long el2_switch,unsigned long entry, +unsigned long arg0,unsigned long arg1) +{ +typeof (__cpu_soft_restart) *restart; + restart = (void *)virt_to_phys (__cpu_soft_restart); +restart (el2_switch, entry, arg0, arg1); +unreachable (); +}#endif
arch/arm64/kernel/machine_kexec.c +1 -2
@@ -210,8 +210,7 @@ void machine_kexec(struct kimage *kimage)* relocation is complete. */ -cpu_soft_restart (virt_to_phys (cpu_reset), - in_crash_kexec ?0 :is_hyp_mode_available (), +cpu_soft_restart (in_crash_kexec ?0 :is_hyp_mode_available (), reboot_code_buffer_phys, kimage->head, kimage_start);BUG ();/* Should never get here. */
Simpler, more readable, no pointless register shifting, smaller idmap footprint.
Thanks,
M.
Signature · 6 lines
Without deviation from the norm, progress is not possible. _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 05/16] arm64: Add back cpu_reset routines
Quoted 2 lines · Marc Zyngier
All that can be solved in C, and mostly at compile time. Using an assembler trampoline is complicating things for no good reason:
I added this into my kexec-v12.1 branch. After some more testing I'll post a v13 to the list.
-Geoff
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 05/16] arm64: Add back cpu_reset routines
Quoted 7 lines · Geoff Levand
Quoted 2 lines · Marc Zyngier
All that can be solved in C, and mostly at compile time. Using an assembler trampoline is complicating things for no good reason:
I added this into my kexec-v12.1 branch. After some more testing I'll post a v13 to the list.
You may well need some notrace annotations if you want to run C code here. Can you give it a spin with things like ftrace and kprobes?
Will
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 05/16] arm64: Add back cpu_reset routines
Hi Will,
Quoted 11 lines · Will Deacon
Quoted 7 lines · Geoff Levand
Quoted 2 lines · Marc Zyngier
All that can be solved in C, and mostly at compile time. Using an assembler trampoline is complicating things for no good reason:
I added this into my kexec-v12.1 branch. After some more testing I'll post a v13 to the list.
You may well need some notrace annotations if you want to run C code here. Can you give it a spin with things like ftrace and kprobes?
I tested ftrace with and without the notrace annotation on cpu_soft_restart() and could do kexec re-boots either way.
When cpu_soft_restart() is called we've only done setup_mm_for_reboot(), so I would think C code should still be OK.
-Geoff
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 05/16] arm64: Add back cpu_reset routines
Quoted 10 lines · Geoff Levand
Quoted 9 lines · Will Deacon
Quoted 7 lines · Geoff Levand
Quoted 2 lines · Marc Zyngier
All that can be solved in C, and mostly at compile time. Using an assembler trampoline is complicating things for no good reason:
I added this into my kexec-v12.1 branch. After some more testing I'll post a v13 to the list.
Quoted 8 lines
Quoted 2 lines
You may well need some notrace annotations if you want to run C code here. Can you give it a spin with things like ftrace and kprobes?
I tested ftrace with and without the notrace annotation on
cpu_soft_restart()and could do kexec re-boots either way.When
cpu_soft_restart()is called we've only donesetup_mm_for_reboot(), so I would think C code should still be OK.
Yeah, you're right. I mistakenly thought that cpu_soft_restart was running at the idmap address as opposed to the kernel virtual address.
Will
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 05/16] arm64: Add back cpu_reset routines
In reply to Marc ZyngierHi Marc,
Quoted 10 lines · Marc Zyngier
Review
Quoted 5 lines · Geoff Levand
(quoted diff) +5
-0+#define SCTLR_ELx_I> > > > (1 << 12) +#define SCTLR_ELx_SA> > > > (1 << 3) +#define SCTLR_ELx_C> > > > (1 << 2) +#define SCTLR_ELx_A> > > > (1 << 1) +#define SCTLR_ELx_M> > > > 1
kvm_arm.halready has some of these defined, andsysreg.hhas another bunch. Can you please consolidate this into a single place (probablysysreg.h)?
OK, I'll look into it.
Quoted 9 lines
arch/arm64/kernel/cpu-reset.S line 28
Quoted 7 lines
arch/arm64/kernel/cpu-reset.S new file +7
-0+#define SCTLR_ELx_FLAGS> > > > (SCTLR_ELx_M | SCTLR_ELx_A | SCTLR_ELx_C |> > > > \ +> > > > > > > > SCTLR_ELx_SA | SCTLR_ELx_I) + +.text +.pushsection .idmap.text, "ax" + +.align 5Why is this .align directive required?
In the original proc.S file, this was to align the cpu_reset routine. It has been there since the initial commit of the proc.S. I don't think this alignment is needed, but maybe there is some reason it is there that I am not aware of?
Quoted 57 lines
Quoted 10 lines
arch/arm64/kernel/cpu-reset.S new file +10
-0+ +/* + * cpu_soft_restart(cpu_reset, el2_switch, entry, arg0, arg1) - Perform a cpu + * soft reset. + * + * @cpu_reset: Physical address of the cpu_reset routine. + * @el2_switch: Flag to indicate a swich to EL2 is needed, passed to cpu_reset. + * @entry: Location to jump to for soft reset, passed to cpu_reset. + * arg0: First argument passed to @entry. + * arg1: Second argument passed to @entry.arch/arm64/kernel/cpu-reset.S line 48
Quoted 10 lines
arch/arm64/kernel/cpu-reset.S new file +10
-0+ */ + +ENTRY(cpu_soft_restart) +> > > > mov> > > > x18, x0> > > > > > > > > > // cpu_reset +> > > > mov> > > > x0, x1> > > > > > > > > > // el2_switch +> > > > mov> > > > x1, x2> > > > > > > > > > // entry +> > > > mov> > > > x2, x3> > > > > > > > > > // arg0 +> > > > mov> > > > x3, x4> > > > > > > > > > // arg1 +> > > > ret> > > > x18 +ENDPROC(cpu_soft_restart)Grepping through the tree, I can only find a single use of
cpu_soft_restart, withcpu_resetas its first parameter.Why do we need this indirection? Having
void
cpu_soft_restart(el2_switch, entry, arg0, arg1);should be enough...
Quoted 27 lines
arch/arm64/kernel/cpu-reset.S new file +27
-0+/* + * cpu_reset(el2_switch, entry, arg0, arg1) - Helper for cpu_soft_restart. + * + * @el2_switch: Flag to indicate a swich to EL2 is needed. + * @entry: Location to jump to for soft reset. + * arg0: First argument passed to @entry. + * arg1: Second argument passed to @entry. + * + * Put the CPU into the same state as it would be if it had been reset, and + * branch to what would be the reset vector. It must be executed with the + * flat identity mapping. + */ + +ENTRY(cpu_reset) +> > > > /* Clear sctlr_el1 flags. */ +> > > > mrs> > > > x12, sctlr_el1 +> > > > ldr> > > > x13, =SCTLR_ELx_FLAGS +> > > > bic> > > > x12, x12, x13 +> > > > msr> > > > sctlr_el1, x12 +> > > > isb + +> > > > cbz> > > > x0, 1f> > > > > > > > > > // el2_switch? + +> > > > mov> > > > x0, x1> > > > > > > > > > // entry +> > > > mov> > > > x1, x2> > > > > > > > > > // arg0 +> > > > mov> > > > x2, x3> > > > > > > > > > // arg1 +> > > > hvc> > > > #HVC_CALL_FUNC> > > > > > > > // no return
Quoted 8 lines
Review
Quoted 6 lines
(quoted diff) +6
-0+ +1:> > > > mov> > > > x18, x1> > > > > > > > > > // entry +> > > > mov> > > > x0, x2> > > > > > > > > > // arg0 +> > > > mov> > > > x1, x3> > > > > > > > > > // arg1 +> > > > ret> > > > x18 +ENDPROC(cpu_reset)And this can be an actual part of
cpu_soft_reset.
Sure, I'll look at your proposal from your reply to Pratyush.
Quoted 13 lines
Quoted 13 lines
(quoted diff) +2
-0+ +.popsectionarch/arm64/kernel/cpu-reset.h new file +5
-0@@ -0,0 +1,22 @@ +/* +* cpu reset routines +* +* Copyright (C) 2015 Huawei Futurewei Technologies. +*
Quoted 8 lines
arch/arm64/kernel/cpu-reset.h line 11
Quoted 6 lines
arch/arm64/kernel/cpu-reset.h new file +6
-0+* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License version 2 as +* published by the Free Software Foundation. +*/ + +#if !defined(_ARM64_CPU_RESET_H)The usual idiom is "#ifndef
__BLAH_H".
Sure, I can change that. Grepping the arm64 sources shows one other use of this style, in a kvm header...
Quoted 6 lines
Quoted 6 lines
arch/arm64/kernel/cpu-reset.h new file +6
-0+#define _ARM64_CPU_RESET_H + +#include + +void __attribute__((noreturn)) cpu_reset(unsigned long el2_switch, +> > > > unsigned long entry, unsigned long arg0, unsigned long arg1);
Quoted 8 lines
arch/arm64/kernel/cpu-reset.h line 22
Quoted 6 lines
arch/arm64/kernel/cpu-reset.h new file +5
-0+void __attribute__((noreturn))cpu_soft_restart (phys_addr_t cpu_reset, +> > > > unsigned long el2_switch, unsigned long entry, unsigned long arg0, +> > > > unsigned long arg1); + +#endif There is a
__noreturnattribute already defined.
I'll switch these over.
Thanks for the review.
-Geoff
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 05/16] arm64: Add back cpu_reset routines
On 30/11/15 20:03, Geoff Levand wrote:
Hi Geoff,
Quoted 8 lines
Quoted 3 lines
Quoted 1 line
+.align 5
Why is this .align directive required?
In the original
proc.Sfile, this was to align thecpu_resetroutine. It has been there since the initial commit of theproc.S. I don't think this alignment is needed, but maybe there is some reason it is there that I am not aware of?
None that I can think of, so just drop it - we'll find out pretty quickly if there was an actual requirement.
Thanks,
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 6/16 Revert "arm64: mm: remove unused cpu_set_idmap_tcr_t0sz function"
In reply to Geoff LevandThis reverts commit c51e97d89e526368eb697f87cd4d391b9e19f369.
Add back the cpu_set_idmap_tcr_t0sz function needed by setup_mmu_for_reboot.
--- arch/arm64/include/asm/mmu_context.h | 35 +++++++++++++++++++++++------------ 1 file changed, 23 insertions(+), 12 deletions(-)
arch/arm64/include/asm/mmu_context.h +23 -12
@@ -70,23 +70,34 @@ static inline bool __cpu_uses_extended_idmap(void) unlikely(idmap_t0sz != TCR_T0SZ(VA_BITS))); } +static inline void __cpu_set_tcr_t0sz (u64 t0sz) +{ +unsigned long tcr; + +if (__cpu_uses_extended_idmap ()) + asmvolatile ( +" mrs %0, tcr_el1 ;" +" bfi %0, %1, %2, %3 ;" +" msr tcr_el1, %0 ;" +" isb" + :"=&r" (tcr) + :"r" (t0sz),"I" (TCR_T0SZ_OFFSET ),"I" (TCR_TxSZ_WIDTH)); +} + +/* +* Set TCR.T0SZ to the value appropriate for activating the identity map. +*/ +static inline void cpu_set_idmap_tcr_t0sz (void ) +{ +__cpu_set_tcr_t0sz (idmap_t0sz); +} +/* * Set TCR.T0SZ to its default value (based on VA_BITS) */ static inline void cpu_set_default_tcr_t0sz (void ) { -unsigned long tcr; - -if (!__cpu_uses_extended_idmap ()) -return ; - - asmvolatile ( -" mrs %0, tcr_el1 ;" -" bfi %0, %1, %2, %3 ;" -" msr tcr_el1, %0 ;" -" isb" - :"=&r" (tcr) - :"r" (TCR_T0SZ (VA_BITS )),"I" (TCR_T0SZ_OFFSET ),"I" (TCR_TxSZ_WIDTH)); +__cpu_set_tcr_t0sz (TCR_T0SZ (VA_BITS )); }/*
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 7/16 Revert "arm64: remove dead code"
In reply to Geoff LevandThis reverts commit b08d4640a3dca68670fc5af2fe9205b395a02388.
Add back the setup_mm_for_reboot() needed for kexec.
--- arch/arm64/include/asm/mmu.h | 1 + arch/arm64/mm/mmu.c | 11 +++++++++++ 2 files changed, 12 insertions(+)
arch/arm64/include/asm/mmu.h +1 -0
@@ -29,6 +29,7 @@ typedef struct {#define ASID (mm) ((mm)->context.id.counter & 0xffff)extern void paging_init (void ); +extern void setup_mm_for_reboot (void );extern void __iomem *early_io_map (phys_addr_t phys,unsigned long virt);extern void init_mem_pgprot (void ); extern void create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys,
arch/arm64/mm/mmu.c +11 -0
@@ -515,6 +515,17 @@ void __init paging_init(void) }/* +* Enable the identity mapping to allow the MMU disabling. +*/ +void setup_mm_for_reboot (void ) +{ +cpu_set_reserved_ttbr0 (); +flush_tlb_all (); +cpu_set_idmap_tcr_t0sz (); +cpu_switch_mm (idmap_pg_dir, &init_mm); +} + +/* * Check whether a kernel address is valid (derived from arch/x86/). */ int kern_addr_valid (unsigned long addr)
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 8/16 arm64/kexec: Add core kexec support
In reply to Geoff LevandAdd three new files, kexec.h, machine_kexec.c and relocate_kernel.S to the arm64 architecture that add support for the kexec re-boot mechanism (CONFIG_KEXEC) on arm64 platforms.
--- arch/arm64/Kconfig | 10 +++ arch/arm64/include/asm/kexec.h | 48 ++++++++++++ arch/arm64/kernel/Makefile | 2 + arch/arm64/kernel/machine_kexec.c | 152 ++++++++++++++++++++++++++++++++++++ arch/arm64/kernel/relocate_kernel.S | 131 +++++++++++++++++++++++++++++++ include/uapi/linux/kexec.h | 1 + 6 files changed, 344 insertions(+) create mode 100644 arch/arm64/include/asm/kexec.h create mode 100644 arch/arm64/kernel/machine_kexec.c create mode 100644 arch/arm64/kernel/relocate_kernel.S
arch/arm64/Kconfig +10 -0
@@ -535,6 +535,16 @@ config SECCOMP and the task is only allowed to execute a few safe syscalls defined by each seccomp mode. +config KEXEC + depends on PM_SLEEP_SMP + select KEXEC_CORE + bool "kexec system call" + ---help--- + kexec is a system call that implements the ability to shutdown your + current kernel, and to start another kernel. It is like a reboot + but it is independent of the system firmware. And like a reboot + you can start any kernel with it, not just Linux. + config XEN_DOM0 def_bool y depends on XEN
arch/arm64/include/asm/kexec.h new file +48 -0
48 lines not on this page · raw message
arch/arm64/kernel/Makefile +2 -0
@@ -41,6 +41,8 @@ arm64-obj-$(CONFIG_EFI) += efi.o efi-entry.stub.o arm64-obj-$(CONFIG_PCI) += pci.o arm64-obj-$(CONFIG_ARMV8_DEPRECATED) += armv8_deprecated.o arm64-obj-$(CONFIG_ACPI) += acpi.o +arm64-obj-$(CONFIG_KEXEC) += machine_kexec.o relocate_kernel.o \ + cpu-reset.o obj-y += $(arm64-obj-y) vdso/ obj-m += $(arm64-obj-m)
arch/arm64/kernel/machine_kexec.c new file +152 -0
152 lines not on this page · raw message
arch/arm64/kernel/relocate_kernel.S new file +131 -0
131 lines not on this page · raw message
include/uapi/linux/kexec.h +1 -0
@@ -39,6 +39,7 @@#define KEXEC_ARCH_SH (42 << 16)#define KEXEC_ARCH_MIPS_LE (10 << 16)#define KEXEC_ARCH_MIPS ( 8 << 16) +#define KEXEC_ARCH_ARM64 (183 << 16)/* The artificial cap on the number of segments passed to kexec_load. */ #define KEXEC_SEGMENT_MAX 16
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Hi Geoff,
arch/arm64/kernel/relocate_kernel.S line 65
Quoted 6 lines · Geoff Levand
arch/arm64/kernel/relocate_kernel.S new file +6
-0+ /* Test the entry flags. */ +.Ltest_source: + tbz x18, IND_SOURCE_BIT, .Ltest_indirection + + mov x20, x13 /* x20 = copy dest */ + mov x21, x12 /* x21 = copy src */
Till v10 we had here invalidation for relocated destination page to PoC. I could not understand, why it was removed. Removing that piece of code breaks kexec booting with mustang. I need [1] to kexec boot into second kernel with mustang platform.
[1]
https://github.com/pratyushanand/linux/commit/431e3247391981a1e8b2864a83a5743e8a274cb9
~Pratyush
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Hi,
Quoted 8 lines · Pratyush Anand
Quoted 6 lines · Geoff Levand
+> > > > /* Test the entry flags. */
+.Ltest_source:
+> > > > tbz> > > > x18,IND_SOURCE_BIT, .Ltest_indirection
+
+> > > > mov x20, x13> > > > > > > > > > /* x20 = copy dest */
+> > > > mov x21, x12> > > > > > > > > > /* x21 = copy src */
Quoted 4 lines
Till v10 we had here invalidation for relocated destination page to PoC. I could not understand, why it was removed. Removing that piece of code breaks kexec booting with mustang. I need [1] to kexec boot into second kernel with mustang platform.
We need to flush the new kernel to PoC. The code that was here that was doing that would only be executed when the new kernel needed relocation (the standard kexec case). We also need to flush kernels that do not need relocation (the standard kdump case).
I moved the new kernel flush to kexec_segment_flush(), called unconditionally in machine_kexec() so we can handle both cases with one piece of code.
Have you experienced a problem on mustang with the current version?
-Geoff
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Hi Geoff,
Quoted 17 lines · Geoff Levand
Hi,
Quoted 8 lines · Pratyush Anand
Quoted 6 lines · Geoff Levand
+> > > > /* Test the entry flags. */
+.Ltest_source:
+> > > > tbz> > > > x18,IND_SOURCE_BIT, .Ltest_indirection
+
+> > > > mov x20, x13> > > > > > > > > > /* x20 = copy dest */
+> > > > mov x21, x12> > > > > > > > > > /* x21 = copy src */Quoted 4 lines
Till v10 we had here invalidation for relocated destination page to PoC. I could not understand, why it was removed. Removing that piece of code breaks kexec booting with mustang. I need [1] to kexec boot into second kernel with mustang platform.
Quoted 8 lines
We need to flush the new kernel to PoC. The code that was here that was doing that would only be executed when the new kernel needed relocation (the standard kexec case). We also need to flush kernels that do not need relocation (the standard kdump case).
I moved the new kernel flush to
kexec_segment_flush(), called unconditionally inmachine_kexec()so we can handle both cases with one piece of code.
Yes, I had noticed that. Actually flushing before cache is disabled can always cause heisenbug.
Quoted 2 lines
Have you experienced a problem on mustang with the current version?
Yes, v10 works fine, but I need invalidation fix for both v11 and v12 to work on mustang.
I have not tested vanilla v12 on seattle, but v12 ported on rhelsa needs invalidation fix to work on seattle as well.
~Pratyush
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
RE: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Hi,
I tested with v11 on our platform and I too needed the invalidation fix.
Thanks,
Azriel Samson
Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
-----Original Message-----
From: kexec [mailto:[email protected]] On Behalf Of Pratyush Anand
Sent: Monday, November 30, 2015 7:16 PM
To: Geoff Levand
Cc: Mark Rutland; [email protected]; Catalin Marinas; Will Deacon; AKASHI Takahiro; [email protected]; Dave Young; [email protected]; [email protected]
Subject: Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Hi Geoff,
Quoted 10 lines · Geoff Levand
Hi,
Quoted 6 lines · Pratyush Anand
Quoted 5 lines · Geoff Levand
+> > > > /* Test the entry flags. */
+.Ltest_source:
+> > > > tbz> > > > x18,IND_SOURCE_BIT, .Ltest_indirection
+
+> > > > mov x20, x13> > > > > > > > > > /* x20
= copy dest */
Quoted 1 line
Quoted 1 line
Quoted 1 line
+> > > > mov x21, x12> > > > > > > > > > /* x21
= copy src */
Quoted 6 lines
Quoted 5 lines
Till v10 we had here invalidation for relocated destination page to PoC. I could not understand, why it was removed. Removing that piece of code breaks kexec booting with mustang. I need [1] to kexec boot into second kernel with mustang platform.
Quoted 8 lines
We need to flush the new kernel to PoC. The code that was here that was doing that would only be executed when the new kernel needed relocation (the standard kexec case). We also need to flush kernels that do not need relocation (the standard kdump case).
I moved the new kernel flush to
kexec_segment_flush(), called unconditionally inmachine_kexec()so we can handle both cases with one piece of code.
Yes, I had noticed that. Actually flushing before cache is disabled can always cause heisenbug.
Quoted 2 lines
Have you experienced a problem on mustang with the current version?
Yes, v10 works fine, but I need invalidation fix for both v11 and v12 to work on mustang.
I have not tested vanilla v12 on seattle, but v12 ported on rhelsa needs invalidation fix to work on seattle as well.
~Pratyush
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Azriel and Pratyush,
Quoted 2 lines · Azriel Samson
I tested with v11 on our platform and I too needed the invalidation fix.
I pushed out a kexec-v12.1 branch to my linux-kexec repo [1] with fixes. Could you test it with both kexec reboot and kdump? For kdump, use something like:
CMDLINE += 'crashkernel=64M@2240M'
then
# kexec -d --load-panic /boot/vmlinux.strip --append="1 mem=64M maxcpus=1 reset_devices console=... root=..." # echo c > /proc/sysrq-trigger
We need to test both reboot and kdump since the the way the kexec is done is different for the two.
[1] https://git.kernel.org/cgit/linux/kernel/git/geoff/linux-kexec.git
-Geoff
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
RE: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Hi Geoff,
Quoted 1 line
I pushed out a kexec-v12.1 branch to my linux-kexec repo [1] with fixes.
Could you test it with both kexec reboot and kdump? For kdump, use something like:
Quoted 1 line
CMDLINE += 'crashkernel=64M@2240M'
For kdump, I also require a version of the patch "arm64/kexec: Add support for kexec-lite". Without this, the dump capture kernel cannot find the device tree even on v11 that I tested.
Thanks,
Azriel Samson
Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Hi,
Quoted 8 lines · Azriel Samson
I pushed out a kexec-v12.1 branch to my linux-kexec repo [1] with fixes. Could you test it with both kexec reboot and kdump? For kdump, use something like:
Quoted 1 line
CMDLINE += 'crashkernel=64M@2240M'
For kdump, I also require a version of the patch "arm64/kexec: Add support for kexec-lite". Without this, the dump capture kernel cannot find the device tree even on v11 that I tested.
Without that patch, does the 2nd stage kernel get the correct dtb address in register x0?
Is the problem that the dtb data is corrupt when in the 2nd stage kernel?
Is the dtb data OK when in the 1st stage kernel?
-Geoff
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Hi,
Quoted 6 lines · Geoff Levand
Hi,
Quoted 3 lines · Azriel Samson
For kdump, I also require a version of the patch "arm64/kexec: Add support for kexec-lite". Without this, the dump capture kernel cannot find the device tree even on v11 that I tested.
This was a setup issue on my side. kdump with your v11 patches worked fine without the above patch when I used your version of kexec-tools.
In kexec-tools, I had to skip check_cpu_nodes() since we are using ACPI and the device tree does not contain cpu_info.
Quoted 11 lines
Without that patch, does the 2nd stage kernel get the correct dtb address in register x0?
Is the problem that the dtb data is corrupt when in the 2nd stage kernel?
Is the dtb data OK when in the 1st stage kernel?
-Geoff
Thanks,
Azriel Samson
Qualcomm Innovation Center, Inc.
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Hi Geoff,
Quoted 5 lines · Azriel Samson
This was a setup issue on my side. kdump with your v11 patches worked fine without the above patch when I used your version of kexec-tools.
In kexec-tools, I had to skip
check_cpu_nodes()since we are using ACPI and the device tree does not containcpu_info.
I think, something like following should be squashed into patch "arm64: Add arm64 kexec support".
https://github.com/pratyushanand/kexec-tools/commit/85f403cabb9ae0d7a37d1656a00fb609a9525086
~Pratyush
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Quoted 6 lines · Pratyush Anand
Hi Geoff,
Quoted 3 lines · Azriel Samson
This was a setup issue on my side. kdump with your v11 patches worked fine without the above patch when I used your version of kexec-tools.
Quoted 6 lines
Quoted 2 lines
In kexec-tools, I had to skip
check_cpu_nodes()since we are using ACPI and the device tree does not containcpu_info.I think, something like following should be squashed into patch "arm64: Add arm64 kexec support".
https://github.com/pratyushanand/kexec-tools/commit/85f403cabb9ae0d7a37d1656a00fb609a9525086
I think it better to test for ACPI and then run
any meaningful checks than to just ignore thecheck_cpu_nodes failure.
-Geoff
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
In reply to Geoff LevandHi Geoff,
Quoted 9 lines · Geoff Levand
Azriel and Pratyush,
Quoted 2 lines · Azriel Samson
I tested with v11 on our platform and I too needed the invalidation fix.
I pushed out a kexec-v12.1 branch to my linux-kexec repo [1] with fixes. Could you test it with both kexec reboot and kdump? For kdump, use something like:
I tested kexec-v12.1 with mustang and it works for both kexec and kdump.
There had been few updates with Akashi, which are not there in kexec-v12.1. Please take his updates and merge them in v13.
~Pratyush
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Mark Rutland
reviewerRe: [PATCH v12 08/16] arm64/kexec: Add core kexec support
In reply to Pratyush AnandQuoted 25 lines · Pratyush Anand
Hi Geoff,
Quoted 22 lines · Geoff Levand
Hi,
Quoted 8 lines · Pratyush Anand
Quoted 6 lines · Geoff Levand
+> > > > /* Test the entry flags. */
+.Ltest_source:
+> > > > tbz> > > > x18,IND_SOURCE_BIT, .Ltest_indirection
+
+> > > > mov x20, x13> > > > > > > > > > /* x20 = copy dest */
+> > > > mov x21, x12> > > > > > > > > > /* x21 = copy src */Quoted 4 lines
Till v10 we had here invalidation for relocated destination page to PoC. I could not understand, why it was removed. Removing that piece of code breaks kexec booting with mustang. I need [1] to kexec boot into second kernel with mustang platform.
We need to flush the new kernel to PoC. The code that was here that was doing that would only be executed when the new kernel needed relocation (the standard kexec case). We also need to flush kernels that do not need relocation (the standard kdump case).
Quoted 6 lines
Quoted 3 lines
I moved the new kernel flush to
kexec_segment_flush(), called unconditionally inmachine_kexec()so we can handle both cases with one piece of code.Yes, I had noticed that. Actually flushing before cache is disabled can always cause heisenbug.
Please use "clean", "invalidate", or "clean+invalidate" rather than "flush". The latter is ambiguous and misleading.
You can validly perform maintenance while the cache may allocate for a region of memory, it's just that afterwards the cache may hold a clean entries for that region.
You can clean/clean+invalidate to push data to the PoC, or you can invalidate/clean+invalidate to ensure that no asynchronous writebacks occur later (so long as you do not make a cacheable write to said cacheline).
The only thing that you cannot guarantee is that there is not some clean cacheline allocated for a region of memory to which cacheable accesses may be performed.
Note that the kernel only requires its Image to be clean to the PoC. So long as this is true, we know that there will not be asynchrnoous writebacks, and can invalidate as necessary as part of the boot process.
Thanks,
Mark.
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Hi Mark,
Quoted 9 lines · Mark Rutland
You can validly perform maintenance while the cache may allocate for a region of memory, it's just that afterwards the cache may hold a clean entries for that region.
You can clean/clean+invalidate to push data to the PoC, or you can invalidate/clean+invalidate to ensure that no asynchronous writebacks occur later (so long as you do not make a cacheable write to said cacheline).
Quoted 7 lines
The only thing that you cannot guarantee is that there is not some clean cacheline allocated for a region of memory to which cacheable accesses may be performed.
Note that the kernel only requires its Image to be clean to the PoC. So long as this is true, we know that there will not be asynchrnoous writebacks, and can invalidate as necessary as part of the boot process.
In v10, which worked for Mustang and Qualcomm, we had:
clean+invalidate to PoC all source pages
disable d-cache
loop {
invalidate to PoC destination page
copy page source->destination
}
enter new image
In v11 I changed this, and it did not work for those platforms:
clean+invalidate to PoC all source pages
clean+invalidate to PoC all destination pages
disable d-cache
loop {
copy page source->destination }
enter new image
Based on your comments above I would think both should work OK.
-Geoff
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Mark Rutland
reviewerRe: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Quoted 19 lines · Geoff Levand
Hi Mark,
Quoted 16 lines · Mark Rutland
You can validly perform maintenance while the cache may allocate for a region of memory, it's just that afterwards the cache may hold a clean entries for that region.
You can clean/clean+invalidate to push data to the PoC, or you can invalidate/clean+invalidate to ensure that no asynchronous writebacks occur later (so long as you do not make a cacheable write to said cacheline).
The only thing that you cannot guarantee is that there is not some clean cacheline allocated for a region of memory to which cacheable accesses may be performed.
Note that the kernel only requires its Image to be clean to the PoC. So long as this is true, we know that there will not be asynchrnoous writebacks, and can invalidate as necessary as part of the boot process.
I've realised that my wording here is somewhat confusing.
When I say "clean to the PoC", I mean that for a given PA the cache can either:
* Not hold an entry (i.e. be invalid).
* Hold a clean entry, where the data is identical to that at the PoC.
Quoted 15 lines
In v10, which worked for Mustang and Qualcomm, we had:
clean+invalidate to PoC all source pages
disable d-cache
loop {
invalidate to PoC destination page
copy pagesource->destination
}
enter new imageIn v11 I changed this, and it did not work for those platforms:
clean+invalidate to PoC all source pages
clean+invalidate to PoC all destination pages
disable d-cache
Quoted 6 lines
loop { copy page source->destination}
enter new imageBased on your comments above I would think both should work OK.
No.
In the latter case, clean lines can be allocated before cacheable data accesses are inhibited. So stale clean lines can shadow the data copied via non-cacheable accesses.
While the cache is clean, it isn't clean to the PoC.
Thanks,
Mark.
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
In reply to Geoff LevandQuoted 47 lines · Geoff Levand
Add three new files,
kexec.h,machine_kexec.candrelocate_kernel.Sto the arm64 architecture that add support for the kexec re-boot mechanism (CONFIG_KEXEC) on arm64 platforms.Signed-off-by Geoff Levand <[email protected]>--- arch/arm64/Kconfig | 10 +++ arch/arm64/include/asm/kexec.h | 48 ++++++++++++ arch/arm64/kernel/Makefile | 2 + arch/arm64/kernel/machine_kexec.c | 152 ++++++++++++++++++++++++++++++++++++ arch/arm64/kernel/relocate_kernel.S | 131 +++++++++++++++++++++++++++++++ include/uapi/linux/kexec.h | 1 + 6 files changed, 344 insertions(+) create mode 100644 arch/arm64/include/asm/kexec.h create mode 100644 arch/arm64/kernel/machine_kexec.c create mode 100644 arch/arm64/kernel/relocate_kernel.Sarch/arm64/Kconfig +10
-0@@ -535,6 +535,16 @@ config SECCOMP and the task is only allowed to execute a few safe syscalls defined by each seccomp mode. +config KEXEC + depends on PM_SLEEP_SMP + select KEXEC_CORE + bool "kexec system call" + ---help--- + kexec is a system call that implements the ability to shutdown your + current kernel, and to start another kernel. It is like a reboot + but it is independent of the system firmware. And like a reboot + you can start any kernel with it, not just Linux. + config XEN_DOM0 def_bool y depends on XENarch/arm64/include/asm/kexec.h new file +3
-0@@ -0,0 +1,48 @@ +/* +* kexec for arm64 +*
arch/arm64/include/asm/kexec.h line 13
Quoted 10 lines
arch/arm64/include/asm/kexec.h new file +10
-0+* Copyright (C) Linaro. +* Copyright (C) Huawei Futurewei Technologies. +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License version 2 as +* published by the Free Software Foundation. +*/ + +#if !defined(_ARM64_KEXEC_H) +#define _ARM64_KEXEC_H
Please keep to the style used elsewhere in the arch headers.
Quoted 4 lines
arch/arm64/include/asm/kexec.h new file +4
-0+ +/* Maximum physical address we can use pages from */ + +#define KEXEC_SOURCE_MEMORY_LIMIT (-1UL)
arch/arm64/include/asm/kexec.h line 27
Quoted 10 lines
arch/arm64/include/asm/kexec.h new file +10
-0+ +/* Maximum address we can reach in physical address mode */ + +#define KEXEC_DESTINATION_MEMORY_LIMIT (-1UL) + +/* Maximum address we can use for the control code buffer */ + +#define KEXEC_CONTROL_MEMORY_LIMIT (-1UL) + +#define KEXEC_CONTROL_PAGE_SIZE 4096
Does this work on kernels configured with 64k pages? It looks like the kexec core code will end up using order-0 pages, so I worry that we'll actually put down 64k and potentially confuse a 4k crash kernel, for example.
arch/arm64/include/asm/kexec.h line 31
Quoted 3 lines
arch/arm64/include/asm/kexec.h new file +3
-0+#define KEXEC_ARCH KEXEC_ARCH_ARM64 + +#if !defined(__ASSEMBLY__)
#ifndef
Quoted 111 lines
arch/arm64/kernel/machine_kexec.c new file +105
-0@@ -0,0 +1,152 @@ +/* +* kexec for arm64 +* +* Copyright (C) Linaro. +* Copyright (C) Huawei Futurewei Technologies. +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License version 2 as +* published by the Free Software Foundation. +*/ + +#include <linux/kexec.h> +#include <linux/of_fdt.h> +#include <linux/slab.h> +#include <linux/uaccess.h> + +#include <asm/cacheflush.h> +#include <asm/system_misc.h> + +#include "cpu-reset.h" + +/* Global variables for the arm64_relocate_new_kernel routine. */ +extern const unsigned char arm64_relocate_new_kernel[]; +extern const unsigned long arm64_relocate_new_kernel_size; + +static unsigned long kimage_start; + +void machine_kexec_cleanup (struct kimage *kimage) +{ +/* Empty routine needed to avoid build errors. */ +} + +/** +* machine_kexec_prepare - Prepare for a kexec reboot. +* +* Called from the core kexec code when a kernel image is loaded. +*/ +int machine_kexec_prepare (struct kimage *kimage) +{ + kimage_start = kimage->start; +return 0 ; +} + +/** +* kexec_list_flush - Helper to flush the kimage list to PoC. +*/ +static void kexec_list_flush (unsigned long kimage_head) +{ +unsigned long *entry; + +for (entry = &kimage_head; ; entry++) { +unsigned int flag = *entry &IND_FLAGS ; +void *addr =phys_to_virt (*entry &PAGE_MASK ); + +switch (flag) { +case IND_INDIRECTION : + entry = (unsigned long *)addr -1 ; +__flush_dcache_area (addr,PAGE_SIZE ); +break ; +case IND_DESTINATION : +break ; +case IND_SOURCE : +__flush_dcache_area (addr,PAGE_SIZE ); +break ; +case IND_DONE : +return ; +default : +BUG (); + } + } +} + +/** +* kexec_segment_flush - Helper to flush the kimage segments to PoC. +*/ +static void kexec_segment_flush (const struct kimage *kimage) +{ +unsigned long i; + +pr_devel ("%s:\n" , __func__); + +for (i =0 ; i < kimage->nr_segments; i++) { +pr_devel (" segment[%lu]: %016lx - %016lx, %lx bytes, %lu pages\n" , + i, + kimage->segment[i].mem, + kimage->segment[i].mem + kimage->segment[i].memsz, + kimage->segment[i].memsz, + kimage->segment[i].memsz /PAGE_SIZE ); + +__flush_dcache_area (phys_to_virt (kimage->segment[i].mem), + kimage->segment[i].memsz); + } +} + +/** +* machine_kexec - Do the kexec reboot. +* +* Called from the core kexec code for a sys_reboot with LINUX_REBOOT_CMD_KEXEC. +*/ +void machine_kexec (struct kimage *kimage) +{ +phys_addr_t reboot_code_buffer_phys; +void *reboot_code_buffer; + +BUG_ON (num_online_cpus () >1 );
arch/arm64/kernel/machine_kexec.c line 115
Quoted 10 lines
arch/arm64/kernel/machine_kexec.c new file +10
-0+ + reboot_code_buffer_phys =page_to_phys (kimage->control_code_page); + reboot_code_buffer =phys_to_virt (reboot_code_buffer_phys); + +/* +* Copy arm64_relocate_new_kernel to the reboot_code_buffer for use +* after the kernel is shut down. +*/ +memcpy (reboot_code_buffer, arm64_relocate_new_kernel, + arm64_relocate_new_kernel_size);
At which point does the I-cache get invalidated for this?
Quoted 4 lines
arch/arm64/kernel/machine_kexec.c new file +4
-0+ +/* Flush the reboot_code_buffer in preparation for its execution. */ +__flush_dcache_area (reboot_code_buffer, arm64_relocate_new_kernel_size); +
arch/arm64/kernel/machine_kexec.c line 129
Quoted 10 lines
arch/arm64/kernel/machine_kexec.c new file +10
-0+/* Flush the new image. */ +kexec_segment_flush (kimage); + +/* Flush the kimage list. */ +kexec_list_flush (kimage->head); + +pr_info ("Bye!\n" ); + +/* Disable all DAIF exceptions. */ + asmvolatile ("msr daifset, #0xf" : : :"memory" );
Can we not use our helpers for this?
Quoted 56 lines
arch/arm64/kernel/machine_kexec.c new file +23
-0+ +setup_mm_for_reboot (); + +/* +* cpu_soft_restart will shutdown the MMU, disable data caches, then +* transfer control to the reboot_code_buffer which contains a copy of +* the arm64_relocate_new_kernel routine. arm64_relocate_new_kernel +* uses physical addressing to relocate the new image to its final +* position and transfers control to the image entry point when the +* relocation is complete. +*/ + +cpu_soft_restart (virt_to_phys (cpu_reset), +is_hyp_mode_available (), + reboot_code_buffer_phys, kimage->head, kimage_start); + +BUG ();/* Should never get here. */ +} + +void machine_crash_shutdown (struct pt_regs *regs) +{ +/* Empty routine needed to avoid build errors. */ +}arch/arm64/kernel/relocate_kernel.S new file +27
-0@@ -0,0 +1,131 @@ +/* + * kexec for arm64 + * + * Copyright (C) Linaro. + * Copyright (C) Huawei Futurewei Technologies. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 as + * published by the Free Software Foundation. + */ + +#include <linux/kexec.h> + +#include <asm/assembler.h> +#include <asm/kexec.h> +#include <asm/kvm_arm.h> +#include <asm/page.h> + +/* + * arm64_relocate_new_kernel - Put a 2nd stage image in place and boot it. + * + * The memory that the old kernel occupies may be overwritten when coping the + * new image to its final location. To assure that the + * arm64_relocate_new_kernel routine which does that copy is not overwritten, + * all code and data needed by arm64_relocate_new_kernel must be between the + * symbols arm64_relocate_new_kernel and arm64_relocate_new_kernel_end. The + * machine_kexec() routine will copy arm64_relocate_new_kernel to the kexec
arch/arm64/kernel/relocate_kernel.S line 37
Quoted 10 lines
arch/arm64/kernel/relocate_kernel.S new file +10
-0+ * control_code_page, a special page which has been set up to be preserved + * during the copy operation. + */ +.globl arm64_relocate_new_kernel +arm64_relocate_new_kernel: + + /* Setup the list loop variables. */ + mov x18, x0 /* x18 = kimage_head */ + mov x17, x1 /* x17 = kimage_start */ + dcache_line_size x16, x0 /* x16 = dcache line size */
Why is this needed?
arch/arm64/kernel/relocate_kernel.S line 47
Quoted 10 lines
arch/arm64/kernel/relocate_kernel.S new file +10
-0+ mov x15, xzr /* x15 = segment start */ + mov x14, xzr /* x14 = entry ptr */ + mov x13, xzr /* x13 = copy dest */ + + /* Clear the sctlr_el2 flags. */ + mrs x0, CurrentEL + cmp x0, #CurrentEL_EL2 + b.ne 1f + mrs x0, sctlr_el2 + ldr x1, =SCTLR_EL2_FLAGS
If we're using literal pools, we probably want a .ltorg directive somewhere.
Quoted 8 lines
arch/arm64/kernel/relocate_kernel.S new file +8
-0+ bic x0, x0, x1 + msr sctlr_el2, x0 + isb +1: + + /* Check if the new image needs relocation. */ + cbz x18, .Ldone + tbnz x18, IND_DONE_BIT, .Ldone
arch/arm64/kernel/relocate_kernel.S line 65
Quoted 10 lines
arch/arm64/kernel/relocate_kernel.S new file +10
-0+ +.Lloop: + and x12, x18, PAGE_MASK /* x12 = addr */ + + /* Test the entry flags. */ +.Ltest_source: + tbz x18, IND_SOURCE_BIT, .Ltest_indirection + + mov x20, x13 /* x20 = copy dest */ + mov x21, x12 /* x21 = copy src */
Weird indentation.
arch/arm64/kernel/relocate_kernel.S line 79
Quoted 13 lines
arch/arm64/kernel/relocate_kernel.S new file +13
-0+ /* Copy page. */ +1: ldp x22, x23, [x21] + ldp x24, x25, [x21, #16] + ldp x26, x27, [x21, #32] + ldp x28, x29, [x21, #48] + add x21, x21, #64 + stnp x22, x23, [x20] + stnp x24, x25, [x20, #16] + stnp x26, x27, [x20, #32] + stnp x28, x29, [x20, #48] + add x20, x20, #64 + tst x21, #(PAGE_SIZE - 1) + b.ne 1b
We should macroise this, to save on duplication of a common routine. You also need to address the caching issues that Mark raised separately.
Quoted 19 lines
arch/arm64/kernel/relocate_kernel.S new file +19
-0+ /* dest += PAGE_SIZE */ + add x13, x13, PAGE_SIZE + b .Lnext + +.Ltest_indirection: + tbz x18, IND_INDIRECTION_BIT, .Ltest_destination + + /* ptr = addr */ + mov x14, x12 + b .Lnext + +.Ltest_destination: + tbz x18, IND_DESTINATION_BIT, .Lnext + + mov x15, x12 + + /* dest = addr */ + mov x13, x12 +
arch/arm64/kernel/relocate_kernel.S line 109
Quoted 10 lines
arch/arm64/kernel/relocate_kernel.S new file +10
-0+.Lnext: + /* entry = *ptr++ */ + ldr x18, [x14], #8 + + /* while (!(entry & DONE)) */ + tbz x18, IND_DONE_BIT, .Lloop + +.Ldone: + dsb sy + ic ialluis
I don't think this needs to be inner-shareable, and these dsbs can probably be non-shareable too.
Quoted 27 lines
arch/arm64/kernel/relocate_kernel.S new file +22
-0+ dsb sy + isb + + /* Start new image. */ + mov x0, xzr + mov x1, xzr + mov x2, xzr + mov x3, xzr + br x17 + +.align 3 /* To keep the 64-bit values below naturally aligned. */ + +.Lcopy_end: +.org KEXEC_CONTROL_PAGE_SIZE + +/* + * arm64_relocate_new_kernel_size - Number of bytes to copy to the + * control_code_page. + */ +.globl arm64_relocate_new_kernel_size +arm64_relocate_new_kernel_size: + .quad .Lcopy_end - arm64_relocate_new_kernelinclude/uapi/linux/kexec.h +0
-0
include/uapi/linux/kexec.h line 42
Quoted 4 lines
include/uapi/linux/kexec.h +1
-0@@ -39,6 +39,7 @@#define KEXEC_ARCH_SH (42 << 16)#define KEXEC_ARCH_MIPS_LE (10 << 16)#define KEXEC_ARCH_MIPS ( 8 << 16) +#define KEXEC_ARCH_ARM64 (183 << 16)
This should probably be called KEXEC_ARCH_AARCH64 for consistency with the ELF machine name.
Will
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Hi Will,
I'll post a v12.4 of this patch that addresses your comments.
Quoted 4 lines · Will Deacon
+#if !defined(_ARM64_KEXEC_H)
Quoted 1 line
+#define
_ARM64_KEXEC_HPlease keep to the style used elsewhere in the arch headers.
OK.
Quoted 7 lines
Review
Quoted 2 lines
(quoted diff) +2
-0+ +#define KEXEC_CONTROL_PAGE_SIZE> > > > 4096Does this work on kernels configured with 64k pages? It looks like the kexec core code will end up using order-0 pages, so I worry that we'll actually put down 64k and potentially confuse a 4k crash kernel, for example.
KEXEC_CONTROL_PAGE_SIZE just tells the core kexec code how big we need control_code_buffer to be. That buffer is only used by the arch code of the first stage kernel. With 64k pages the buffer will be a full page, but we'll only use the first 4k of it.
Quoted 5 lines
arch/arm64/include/asm/kexec.h line 31
Quoted 3 lines
arch/arm64/include/asm/kexec.h new file +3
-0+#define KEXEC_ARCH KEXEC_ARCH_ARM64 + +#if !defined(__ASSEMBLY__)#ifndef
OK.
Quoted 14 lines
Quoted 14 lines
+ *
machine_kexec- Do the kexec reboot.
+ *
+ * Called from the core kexec code for asys_rebootwithLINUX_REBOOT_CMD_KEXEC.
+ */arch/arm64/kernel/machine_kexec.c new file +10
-0+void machine_kexec (struct kimage *kimage) +{ +> > > > phys_addr_t reboot_code_buffer_phys; +> > > > void *reboot_code_buffer; + +> > > > BUG_ON(num_online_cpus() > 1); + +> > > > reboot_code_buffer_phys = page_to_phys(kimage->control_code_page); +> > > > reboot_code_buffer = phys_to_virt(reboot_code_buffer_phys); +
Quoted 8 lines
Quoted 6 lines
+> > > > /*
+> > > > * Copy arm64_relocate_new_kernel to the reboot_code_buffer for use +> > > > * after the kernel is shut down. +> > > > */+> > > > memcpy(reboot_code_buffer,
arm64_relocate_new_kernel,
+> > > > > >arm64_relocate_new_kernel_size);At which point does the I-cache get invalidated for this?
I'll add a call to flush_icache_range() for reboot_code_buffer. I think that should do it.
Quoted 9 lines
Quoted 9 lines
(quoted diff) +9
-0+ +> > > > /* Flush the reboot_code_buffer in preparation for its execution. */ +> > > > __flush_dcache_area(reboot_code_buffer, arm64_relocate_new_kernel_size); + +> > > > /* Flush the new image. */ +> > > > kexec_segment_flush(kimage); + +> > > > /* Flush the kimage list. */ +> > > > kexec_list_flush(kimage->head);
Quoted 8 lines
Quoted 6 lines
(quoted diff) +5
-0+ +> > > > pr_info("Bye!\n"); + +> > > > /* Disable all DAIF exceptions. */ +> > > > asm volatile ("msr > > daifset> > , #0xf" : : :"memory");
Can we not use our helpers for this?
Mark Rutland had commented that calling daifset four times through the different macros took considerable time, and recommended a single call here.
Would you prefer a new macro for irqflags.h, maybe
#define local_daif_disable() asm("msr daifset, #0xf" : : : "memory")?
Quoted 13 lines
Quoted 13 lines
+/*
+ *arm64_relocate_new_kernel- Put a 2nd stage image in place and boot it.
+ *
+ * The memory that the old kernel occupies may be overwritten when coping the
+ * new image to its final location. To assure that the
+ *arm64_relocate_new_kernelroutine which does that copy is not overwritten,
+ * all code and data needed byarm64_relocate_new_kernelmust be between the
+ * symbolsarm64_relocate_new_kernelandarm64_relocate_new_kernel_end. The
+ *machine_kexec()routine will copyarm64_relocate_new_kernelto the kexec
+ *control_code_page, a special page which has been set up to be preserved
+ * during the copy operation.
+ */
+.globlarm64_relocate_new_kernel
Quoted 8 lines
Review
Quoted 6 lines
(quoted diff) +6
-0+arm64_relocate_new_kernel: + +> > > > /* Setup the list loop variables. */ +> > > > mov> > > > x18, x0> > > > > > > > > > /* x18 = kimage_head */ +> > > > mov> > > > x17, x1> > > > > > > > > > /* x17 = kimage_start */ +> > > > dcache_line_size x16, x0> > > > > > /* x16 = dcache line size */Why is this needed?
This was left over from the previous version where we invalidated pages while copying them. I've since added that invalidate back, so this is again needed.
Quoted 12 lines
Quoted 11 lines
+> > > > mov> > > > x15, xzr> > > > > > > > /* x15 = segment start */
+> > > > mov> > > > x14, xzr> > > > > > > > /* x14 = entry ptr */
+> > > > mov> > > > x13, xzr> > > > > > > > /* x13 = copy dest */
+
+> > > > /* Clear thesctlr_el2flags. */
+> > > > mrs> > > > x0, CurrentEL
+> > > > cmp> > > > x0, #CurrentEL_EL2
+> > > > b.ne> > > > 1f
+> > > > mrs> > > > x0,sctlr_el2
+> > > > ldr> > > > x1, =SCTLR_EL2_FLAGS
Quoted 1 line
If we're using literal pools, we probably want a .ltorg directive somewhere.
I've added one in at the end of the arm64_relocate_new_kernel code.
Quoted 12 lines
Quoted 12 lines
+> > > > bic> > > > x0, x0, x1
+> > > > msr> > > >sctlr_el2, x0
+> > > > isb
+1:arch/arm64/kernel/relocate_kernel.S new file +8
-0+ +> > > > /* Check if the new image needs relocation. */ +> > > > cbz> > > > x18, .Ldone +> > > > tbnz> > > > x18, IND_DONE_BIT, .Ldone + +.Lloop: +> > > > and> > > > x12, x18, PAGE_MASK> > > > > > /* x12 = addr */ +
Quoted 8 lines
Quoted 6 lines
+> > > > /* Test the entry flags. */
+.Ltest_source:
+> > > > tbz> > > > x18,IND_SOURCE_BIT, .Ltest_indirection
+
+> > > > mov x20, x13> > > > > > > > > > /* x20 = copy dest */
+> > > > mov x21, x12> > > > > > > > > > /* x21 = copy src */Weird indentation.
Fixed.
Quoted 14 lines
Quoted 5 lines
+> > > > /* Copy page. */
+1:> > > > ldp> > > > x22, x23, [x21]
+> > > > ldp> > > > x24, x25, [x21, #16]
+> > > > ldp> > > > x26, x27, [x21, #32]
+> > > > ldp> > > > x28, x29, [x21, #48]Quoted 8 lines
+> > > > add> > > > x21, x21, #64
+> > > > stnp> > > > x22, x23, [x20]
+> > > > stnp> > > > x24, x25, [x20, #16]
+> > > > stnp> > > > x26, x27, [x20, #32]
+> > > > stnp> > > > x28, x29, [x20, #48]
+> > > > add> > > > x20, x20, #64
+> > > > tst> > > > x21, #(PAGE_SIZE - 1)
+> > > > b.ne> > > > 1b
Quoted 1 line
We should macroise this, to save on duplication of a common routine.
So something like this in assembler.h?
+/*
+ * copy_page - copy src to dest using temp registers t1-t8
+ */
+ .macro copy_page dest:req src:req t1:req t2:req t3:req t4:req t5:req t6:req t7:req t8:req
+1: ldp /t1, /t2, [/src]
+ ldp /t3, /t4, [/src, #16]
+ ldp /t5, /t6, [/src, #32]
+ ldp /t7, /t8, [/src, #48]
+ add /src, /src, #64
+ stnp /t1, /t2, [/dest]
+ stnp /t3, /t4, [/dest, #16]
+ stnp /t5, /t6, [/dest, #32]
+ stnp /t7, /t8, [/dest, #48]
+ add /dest, /dest, #64
+ tst /src, #(PAGE_SIZE - 1)
+ b.ne 1b
+ .endm
Quoted 1 line
You also need to address the caching issues that Mark raised separately.
Cache maintenance has been fixed (reintroduced) in the current code.
Quoted 25 lines
Quoted 25 lines
+> > > > /* dest +=
PAGE_SIZE*/
+> > > > add> > > > x13, x13,PAGE_SIZE
+> > > > b> > > > .Lnext(quoted diff) +21
-0+ +.Ltest_indirection: +> > > > tbz> > > > x18, IND_INDIRECTION_BIT, .Ltest_destination + +> > > > /* ptr = addr */ +> > > > mov> > > > x14, x12 +> > > > b> > > > .Lnext + +.Ltest_destination: +> > > > tbz> > > > x18, IND_DESTINATION_BIT, .Lnext + +> > > > mov> > > > x15, x12 + +> > > > /* dest = addr */ +> > > > mov> > > > x13, x12 + +.Lnext: +> > > > /* entry = *ptr++ */ +> > > > ldr> > > > x18, [x14], #8 + +> > > > /* while (!(entry & DONE)) */
Quoted 8 lines
Quoted 5 lines
+> > > > tbz> > > > x18,
IND_DONE_BIT, .Lloop
+
+.Ldone:
+> > > > dsb> > > > sy
+> > > > ic> > > > ialluisI don't think this needs to be inner-shareable, and these dsbs can probably be non-shareable too.
OK.
Quoted 26 lines
Quoted 26 lines
+> > > > dsb> > > > sy
+> > > > isbarch/arm64/kernel/relocate_kernel.S new file +20
-0+ +> > > > /* Start new image. */ +> > > > mov> > > > x0, xzr +> > > > mov> > > > x1, xzr +> > > > mov> > > > x2, xzr +> > > > mov> > > > x3, xzr +> > > > br> > > > x17 + +.align 3> > > > /* To keep the 64-bit values below naturally aligned. */ + +.Lcopy_end: +.org> > > > KEXEC_CONTROL_PAGE_SIZE + +/* + * arm64_relocate_new_kernel_size - Number of bytes to copy to the + * control_code_page. + */ +.globl arm64_relocate_new_kernel_size +arm64_relocate_new_kernel_size: +> > > > .quad> > > > .Lcopy_end - arm64_relocate_new_kernelinclude/uapi/linux/kexec.h +0
-0
Quoted 8 lines
include/uapi/linux/kexec.h line 42
Quoted 5 lines
include/uapi/linux/kexec.h +1
-0@@ -39,6 +39,7 @@#define KEXEC_ARCH_SH (42 << 16)#define KEXEC_ARCH_MIPS_LE (10 << 16)#define KEXEC_ARCH_MIPS ( 8 << 16) +#define KEXEC_ARCH_ARM64 (183 << 16)This should probably be called
KEXEC_ARCH_AARCH64for consistency with the ELF machine name.
OK.
-Geoff
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Quoted 11 lines · Geoff Levand
Quoted 7 lines · Will Deacon
Quoted 5 lines
+> > > > * Copy arm64_relocate_new_kernel to the reboot_code_buffer for use +> > > > * after the kernel is shut down. +> > > > */+> > > > memcpy(reboot_code_buffer,
arm64_relocate_new_kernel,
+> > > > > >arm64_relocate_new_kernel_size);At which point does the I-cache get invalidated for this?
I'll add a call to
flush_icache_range()forreboot_code_buffer. I think that should do it.
We execute arm64_relocate_new_kernel() code with I-cache disabled. So, do we really need to invalidate I-cache?
~Pratyush
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Quoted 8 lines · Pratyush Anand
Quoted 7 lines · Geoff Levand
Quoted 6 lines · Will Deacon
Quoted 5 lines
+> > > > * Copy arm64_relocate_new_kernel to the reboot_code_buffer for use +> > > > * after the kernel is shut down. +> > > > */+> > > > memcpy(reboot_code_buffer,
arm64_relocate_new_kernel,
+> > > > > >arm64_relocate_new_kernel_size);
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
At which point does the I-cache get invalidated for this?
I'll add a call to
flush_icache_range()forreboot_code_buffer. I think that should do it.We execute
arm64_relocate_new_kernel()code with I-cache disabled. So, do we really need to invalidate I-cache?
I got bitten by this, see Mark's earlier reply[0]:
Quoted 3 lines · Mark Rutland
The SCTLR_ELx.I only affects the attributes that the I-cache uses to fetch with, not whether it is enabled (it cannot be disabled architecturally).
James
[0]
http://lists.infradead.org/pipermail/linux-arm-kernel/2015-October/382101.html
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 08/16] arm64/kexec: Add core kexec support
Quoted 10 lines · James Morse
Quoted 2 lines · Pratyush Anand
We execute
arm64_relocate_new_kernel()code with I-cache disabled. So, do we really need to invalidate I-cache?I got bitten by this, see Mark's earlier reply[0]:
Quoted 3 lines · Mark Rutland
The SCTLR_ELx.I only affects the attributes that the I-cache uses to fetch with, not whether it is enabled (it cannot be disabled architecturally).
Thanks James for pointing to it. I had missed that.
~Pratyush
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
PATCH 1/1 arm64/kexec: Add core kexec support
In reply to Geoff LevandAdd three new files, kexec.h, machine_kexec.c and relocate_kernel.S to the arm64 architecture that add support for the kexec re-boot mechanism (CONFIG_KEXEC) on arm64 platforms.
---
I stuck the copy_page macro in here for review.
-Geoff
arch/arm64/Kconfig | 10 +++ arch/arm64/include/asm/kexec.h | 48 +++++++++++ arch/arm64/kernel/Makefile | 2 + arch/arm64/kernel/machine_kexec.c | 153 ++++++++++++++++++++++++++++++++++++ arch/arm64/kernel/relocate_kernel.S | 131 ++++++++++++++++++++++++++++++ include/uapi/linux/kexec.h | 1 + 6 files changed, 345 insertions(+) create mode 100644 arch/arm64/include/asm/kexec.h create mode 100644 arch/arm64/kernel/machine_kexec.c create mode 100644 arch/arm64/kernel/relocate_kernel.S
arch/arm64/include/asm/assembler.h +19 -0
@@ -24,6 +24,7 @@#define __ASM_ASSEMBLER_H#include <asm/asm-offsets.h> +#include <asm/page.h> #include <asm/pgtable-hwdef.h> #include <asm/ptrace.h> #include <asm/thread_info.h>
@@ -240,6 +241,24 @@ lr .req x30 // link register .endm /* + * copy_page - copy src to dest using temp registers t1-t8 + */ + .macro copy_page dest:req src:req t1:req t2:req t3:req t4:req t5:req t6:req t7:req t8:req +1: ldp \t1, \t2, [\src] + ldp \t3, \t4, [\src, #16] + ldp \t5, \t6, [\src, #32] + ldp \t7, \t8, [\src, #48] + add \src, \src, #64 + stnp \t1, \t2, [\dest] + stnp \t3, \t4, [\dest, #16] + stnp \t5, \t6, [\dest, #32] + stnp \t7, \t8, [\dest, #48] + add \dest, \dest, #64 + tst \src, #(PAGE_SIZE - 1) + b.ne 1b + .endm + +/* * Annotate a function as position independent, i.e., safe to be called before * the kernel virtual mapping is activated. */
arch/arm64/Kconfig +10 -0
@@ -556,6 +556,16 @@ config SECCOMP and the task is only allowed to execute a few safe syscalls defined by each seccomp mode. +config KEXEC + depends on PM_SLEEP_SMP + select KEXEC_CORE + bool "kexec system call" + ---help--- + kexec is a system call that implements the ability to shutdown your + current kernel, and to start another kernel. It is like a reboot + but it is independent of the system firmware. And like a reboot + you can start any kernel with it, not just Linux. + config XEN_DOM0 def_bool y depends on XEN
arch/arm64/include/asm/kexec.h new file +48 -0
48 lines not on this page · raw message
arch/arm64/kernel/Makefile +2 -0
@@ -41,6 +41,8 @@ arm64-obj-$(CONFIG_EFI) += efi.o efi-entry.stub.o arm64-obj-$(CONFIG_PCI) += pci.o arm64-obj-$(CONFIG_ARMV8_DEPRECATED) += armv8_deprecated.o arm64-obj-$(CONFIG_ACPI) += acpi.o +arm64-obj-$(CONFIG_KEXEC) += machine_kexec.o relocate_kernel.o \ + cpu-reset.o obj-y += $(arm64-obj-y) vdso/ obj-m += $(arm64-obj-m)
arch/arm64/kernel/machine_kexec.c new file +153 -0
153 lines not on this page · raw message
arch/arm64/kernel/relocate_kernel.S new file +131 -0
131 lines not on this page · raw message
include/uapi/linux/kexec.h +1 -0
@@ -39,6 +39,7 @@#define KEXEC_ARCH_SH (42 << 16)#define KEXEC_ARCH_MIPS_LE (10 << 16)#define KEXEC_ARCH_MIPS ( 8 << 16) +#define KEXEC_ARCH_AARCH64 (183 << 16)/* The artificial cap on the number of segments passed to kexec_load. */ #define KEXEC_SEGMENT_MAX 16
Signature · 7 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 9/16 arm64/kexec: Add pr_devel output
In reply to Geoff LevandTo aid in debugging kexec problems or when adding new functionality to kexec add a new routine kexec_image_info() and several inline pr_devel statements.
--- arch/arm64/kernel/machine_kexec.c | 63 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 63 insertions(+)
arch/arm64/kernel/machine_kexec.c +63 -0
@@ -25,6 +25,48 @@ extern const unsigned long arm64_relocate_new_kernel_size;static unsigned long kimage_start; +/** +* kexec_is_dtb - Helper routine to check the device tree header signature. +*/ +static bool kexec_is_dtb (const void *dtb) +{ +__be32 magic; + +return get_user (magic, (__be32 *)dtb) ? false : + (be32_to_cpu (magic) ==OF_DT_HEADER ); +} + +/** +* kexec_image_info - For debugging output. +*/ +#define kexec_image_info (_i) _kexec_image_info(__func__, __LINE__, _i) +static void _kexec_image_info (const char *func,int line, +const struct kimage *kimage) +{ +unsigned long i; + +#if !defined(DEBUG ) +return ; +#endif +pr_devel ("%s:%d:\n" , func, line); +pr_devel (" kexec kimage info:\n" ); +pr_devel (" type: %d\n" , kimage->type); +pr_devel (" start: %lx\n" , kimage->start); +pr_devel (" head: %lx\n" , kimage->head); +pr_devel (" nr_segments: %lu\n" , kimage->nr_segments); + +for (i =0 ; i < kimage->nr_segments; i++) { +pr_devel (" segment[%lu]: %016lx - %016lx, %lx bytes, %lu pages%s\n" , + i, + kimage->segment[i].mem, + kimage->segment[i].mem + kimage->segment[i].memsz, + kimage->segment[i].memsz, + kimage->segment[i].memsz /PAGE_SIZE , + (kexec_is_dtb (kimage->segment[i].buf) ? +", dtb segment" :"" )); + } +} +void machine_kexec_cleanup (struct kimage *kimage) {/* Empty routine needed to avoid build errors. */
@@ -38,6 +80,8 @@ void machine_kexec_cleanup(struct kimage *kimage)int machine_kexec_prepare (struct kimage *kimage) { kimage_start = kimage->start; +kexec_image_info (kimage); +return 0 ; }
@@ -107,6 +151,25 @@ void machine_kexec(struct kimage *kimage) reboot_code_buffer_phys =page_to_phys (kimage->control_code_page); reboot_code_buffer =phys_to_virt (reboot_code_buffer_phys); +kexec_image_info (kimage); + +pr_devel ("%s:%d: control_code_page: %p\n" , __func__, __LINE__, + kimage->control_code_page); +pr_devel ("%s:%d: reboot_code_buffer_phys: %pa\n" , __func__, __LINE__, + &reboot_code_buffer_phys); +pr_devel ("%s:%d: reboot_code_buffer: %p\n" , __func__, __LINE__, + reboot_code_buffer); +pr_devel ("%s:%d: relocate_new_kernel: %p\n" , __func__, __LINE__, + arm64_relocate_new_kernel); +pr_devel ("%s:%d: relocate_new_kernel_size: 0x%lx(%lu) bytes\n" , + __func__, __LINE__, arm64_relocate_new_kernel_size, + arm64_relocate_new_kernel_size); + +pr_devel ("%s:%d: kimage_head: %lx\n" , __func__, __LINE__, + kimage->head); +pr_devel ("%s:%d: kimage_start: %lx\n" , __func__, __LINE__, + kimage_start); +/* * Copy arm64_relocate_new_kernel to the reboot_code_buffer for use * after the kernel is shut down.
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 09/16] arm64/kexec: Add pr_devel output
Quoted 2 lines · Geoff Levand
To aid in debugging kexec problems or when adding new functionality to kexec add a new routine
kexec_image_info()and several inlinepr_develstatements.
We don't currently have any pr_devel invocations under arm64. Can you either remove these or change them to pr_debug?
Quoted 12 lines
Signed-off-by Geoff Levand <[email protected]>--- arch/arm64/kernel/machine_kexec.c | 63 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 63 insertions(+)arch/arm64/kernel/machine_kexec.c +0
-0@@ -25,6 +25,48 @@ extern const unsigned long arm64_relocate_new_kernel_size;static unsigned long kimage_start;
arch/arm64/kernel/machine_kexec.c extern const unsigned long arm64_relocate_new_kernel_size; · line 36
Quoted 10 lines
arch/arm64/kernel/machine_kexec.c +9
-0+/** +* kexec_is_dtb - Helper routine to check the device tree header signature. +*/ +static bool kexec_is_dtb (const void *dtb) +{ +__be32 magic; + +return get_user (magic, (__be32 *)dtb) ? false : + (be32_to_cpu (magic) ==OF_DT_HEADER );
Isn't there an fdt32_to_cpu helper for this already? I also can't help but thing this would be more readable if you didn't bother with the ternary if form.
Will
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 09/16] arm64/kexec: Add pr_devel output
Quoted 6 lines · Will Deacon
Quoted 2 lines · Geoff Levand
To aid in debugging kexec problems or when adding new functionality to kexec add a new routine
kexec_image_info()and several inlinepr_develstatements.We don't currently have any
pr_develinvocations under arm64. Can you either remove these or change them topr_debug?
OK.
Quoted 13 lines
Quoted 13 lines
Signed-off-by Geoff Levand <[email protected]>--- arch/arm64/kernel/machine_kexec.c | 63 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 63 insertions(+)arch/arm64/kernel/machine_kexec.c +0
-0@@ -25,6 +25,48 @@ extern const unsigned long arm64_relocate_new_kernel_size;static unsigned long kimage_start;
Quoted 13 lines
arch/arm64/kernel/machine_kexec.c extern const unsigned long arm64_relocate_new_kernel_size;
Quoted 9 lines
arch/arm64/kernel/machine_kexec.c +9
-0+/** +* kexec_is_dtb - Helper routine to check the device tree header signature. +*/ +static bool kexec_is_dtb (const void *dtb) +{ +> > > > __be32 magic; + +> > > > return get_user(magic, (__be32 *)dtb) ? false : +> > > > > > (be32_to_cpu(magic) == OF_DT_HEADER);Isn't there an
fdt32_to_cpuhelper for this already? I also can't help but thing this would be more readable if you didn't bother with the ternary if form.
Sure, I'll clean it up.
-Geoff
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
PATCH 1/1 arm64/kexec: Add pr_debug output
In reply to Geoff LevandTo aid in debugging kexec problems or when adding new functionality to kexec add a new routine kexec_image_info() and several inline pr_debug statements.
--- arch/arm64/kernel/machine_kexec.c | 70 +++++++++++++++++++++++++++++++++++++-- 1 file changed, 68 insertions(+), 2 deletions(-)
arch/arm64/kernel/machine_kexec.c +68 -2
@@ -10,6 +10,7 @@*/ #include <linux/kexec.h> +#include <linux/libfdt_env.h> #include <linux/of_fdt.h> #include <linux/slab.h> #include <linux/uaccess.h>
@@ -25,6 +26,50 @@ extern const unsigned long arm64_relocate_new_kernel_size;static unsigned long kimage_start; +/** +* kexec_is_dtb - Helper routine to check the device tree header signature. +*/ +static bool kexec_is_dtb (const void *dtb) +{ +__be32 magic; + +if (!get_user (magic, (__be32 *)dtb)) +return false; + +return fdt32_to_cpu (magic) ==OF_DT_HEADER ; +} + +/** +* kexec_image_info - For debugging output. +*/ +#define kexec_image_info (_i) _kexec_image_info(__func__, __LINE__, _i) +static void _kexec_image_info (const char *func,int line, +const struct kimage *kimage) +{ +unsigned long i; + +#if !defined(DEBUG ) +return ; +#endif +pr_debug ("%s:%d:\n" , func, line); +pr_debug (" kexec kimage info:\n" ); +pr_debug (" type: %d\n" , kimage->type); +pr_debug (" start: %lx\n" , kimage->start); +pr_debug (" head: %lx\n" , kimage->head); +pr_debug (" nr_segments: %lu\n" , kimage->nr_segments); + +for (i =0 ; i < kimage->nr_segments; i++) { +pr_debug (" segment[%lu]: %016lx - %016lx, %lx bytes, %lu pages%s\n" , + i, + kimage->segment[i].mem, + kimage->segment[i].mem + kimage->segment[i].memsz, + kimage->segment[i].memsz, + kimage->segment[i].memsz /PAGE_SIZE , + (kexec_is_dtb (kimage->segment[i].buf) ? +", dtb segment" :"" )); + } +} +void machine_kexec_cleanup (struct kimage *kimage) {/* Empty routine needed to avoid build errors. */
@@ -38,6 +83,8 @@ void machine_kexec_cleanup(struct kimage *kimage)int machine_kexec_prepare (struct kimage *kimage) { kimage_start = kimage->start; +kexec_image_info (kimage); +return 0 ; }
@@ -77,10 +124,10 @@ static void kexec_segment_flush(const struct kimage *kimage) {unsigned long i; -pr_devel ("%s:\n" , __func__); +pr_debug ("%s:\n" , __func__);for (i =0 ; i < kimage->nr_segments; i++) { - pr_devel(" segment[%lu]: %016lx - %016lx, %lx bytes, %lu pages\n", + pr_debug(" segment[%lu]: %016lx - %016lx, %lx bytes, %lu pages\n", i, kimage->segment[i].mem, kimage->segment[i].mem + kimage->segment[i].memsz,
@@ -107,6 +154,25 @@ void machine_kexec(struct kimage *kimage) reboot_code_buffer_phys =page_to_phys (kimage->control_code_page); reboot_code_buffer =phys_to_virt (reboot_code_buffer_phys); +kexec_image_info (kimage); + +pr_debug ("%s:%d: control_code_page: %p\n" , __func__, __LINE__, + kimage->control_code_page); +pr_debug ("%s:%d: reboot_code_buffer_phys: %pa\n" , __func__, __LINE__, + &reboot_code_buffer_phys); +pr_debug ("%s:%d: reboot_code_buffer: %p\n" , __func__, __LINE__, + reboot_code_buffer); +pr_debug ("%s:%d: relocate_new_kernel: %p\n" , __func__, __LINE__, + arm64_relocate_new_kernel); +pr_debug ("%s:%d: relocate_new_kernel_size: 0x%lx(%lu) bytes\n" , + __func__, __LINE__, arm64_relocate_new_kernel_size, + arm64_relocate_new_kernel_size); + +pr_debug ("%s:%d: kimage_head: %lx\n" , __func__, __LINE__, + kimage->head); +pr_debug ("%s:%d: kimage_start: %lx\n" , __func__, __LINE__, + kimage_start); +/* * Copy arm64_relocate_new_kernel to the reboot_code_buffer for use * after the kernel is shut down.
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 10/16 arm64/kexec: Enable kexec in the arm64 defconfig
In reply to Geoff Levand--- arch/arm64/configs/defconfig | 1 + 1 file changed, 1 insertion(+)
arch/arm64/configs/defconfig +1 -0
@@ -58,6 +58,7 @@ CONFIG_PREEMPT=y CONFIG_KSM=y CONFIG_TRANSPARENT_HUGEPAGE=y CONFIG_CMA=y +CONFIG_KEXEC=y CONFIG_CMDLINE="console=ttyAMA0" # CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS is not set CONFIG_COMPAT=y
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 11/16 arm64: kdump: reserve memory for crash dump kernel
In reply to Geoff LevandFrom: AKASHI Takahiro <[email protected]>
On primary kernel, the memory region used by crash dump kernel must be specified by "crashkernel=" boot parameter. reserve_crashkernel() will allocate and reserve the region for later use.
User space tools will be able to find the region marked as "Crash kernel" in /proc/iomem.
--- arch/arm64/kernel/setup.c | 7 +++++- arch/arm64/mm/init.c | 54 +++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 60 insertions(+), 1 deletion(-)
arch/arm64/kernel/setup.c +6 -1
@@ -31,7 +31,6 @@#include <linux/screen_info.h> #include <linux/init.h> #include <linux/kexec.h> -#include <linux/crash_dump.h> #include <linux/root_dev.h> #include <linux/cpu.h> #include <linux/interrupt.h>
@@ -221,6 +220,12 @@ static void __init request_standard_resources(void) kernel_data.end <= res->end)request_resource (res, &kernel_data); } + +#ifdef CONFIG_KEXEC +/* User space tools will find "Crash kernel" region in /proc/iomem. */ +if (crashk_res.end) +insert_resource (&iomem_resource, &crashk_res); +#endif }#ifdef CONFIG_BLK_DEV_INITRD
arch/arm64/mm/init.c +54 -0
72 lines not on this page · raw message
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 11/16] arm64: kdump: reserve memory for crash dump kernel
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
On primary kernel, the memory region used by crash dump kernel must be specified by "crashkernel=" boot parameter.
reserve_crashkernel()will allocate and reserve the region for later use.User space tools will be able to find the region marked as "Crash kernel" in
/proc/iomem.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm64/kernel/setup.c | 7 +++++- arch/arm64/mm/init.c | 54 +++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 60 insertions(+), 1 deletion(-)arch/arm64/kernel/setup.c +0
-1@@ -31,7 +31,6 @@#include <linux/screen_info.h> #include <linux/init.h> #include <linux/kexec.h> -#include <linux/crash_dump.h> #include <linux/root_dev.h> #include <linux/cpu.h> #include <linux/interrupt.h>
arch/arm64/kernel/setup.c static void __init request_standard_resources(void) · line 228
Quoted 9 lines
arch/arm64/kernel/setup.c +6
-0@@ -221,6 +220,12 @@ static void __init request_standard_resources(void) kernel_data.end <= res->end)request_resource (res, &kernel_data); } + +#ifdef CONFIG_KEXEC +/* User space tools will find "Crash kernel" region in /proc/iomem. */ +if (crashk_res.end) +insert_resource (&iomem_resource, &crashk_res); +#endif
Why can't this bit be moved into reserved_crashkernel, at the end?
Quoted 26 lines
arch/arm64/kernel/setup.c +0
-0}#ifdef CONFIG_BLK_DEV_INITRD arch/arm64/mm/init.c +8
-0@@ -34,6 +34,7 @@#include <linux/dma-contiguous.h> #include <linux/efi.h> #include <linux/swiotlb.h> +#include <linux/kexec.h> #include <asm/fixmap.h> #include <asm/memory.h> @@ -66,6 +67,55 @@ static int __init early_initrd(char *p)early_param ("initrd" , early_initrd);#endif +#ifdef CONFIG_KEXEC +/* +* reserve_crashkernel() - reserves memory for crash kernel +* +* This function reserves memory area given in "crashkernel=" kernel command +* line parameter. The memory reserved is used by dump capture kernel when +* primary kernel is crashing.
arch/arm64/mm/init.c static int __init early_initrd(char *p) · line 86
Quoted 10 lines
arch/arm64/mm/init.c +10
-0+*/ +static void __init reserve_crashkernel (void ) +{ +unsigned long long crash_size =0 , crash_base =0 ; +int ret; + + ret =parse_crashkernel (boot_command_line,memblock_phys_mem_size (), + &crash_size, &crash_base); +if (ret) +return ;
What's the point in ret?
arch/arm64/mm/init.c static int __init early_initrd(char *p) · line 89
Quoted 3 lines
arch/arm64/mm/init.c +3
-0+ +if (crash_base ==0 ) { + crash_base =memblock_alloc (crash_size,1 <<21 );
What is this magic number? How about SZ_2M and a comment mentioning the arm64 boot protocol?
Quoted 4 lines
arch/arm64/mm/init.c +4
-0+if (crash_base ==0 ) { +pr_warn ("Unable to allocate crashkernel (size:%llx)\n" , + crash_size); +return ;
arch/arm64/mm/init.c static int __init early_initrd(char *p) · line 103
Quoted 10 lines
arch/arm64/mm/init.c +10
-0+ } + }else { +/* User specifies base address explicitly. */ +if (!memblock_is_region_memory (crash_base, crash_size) || +memblock_is_region_reserved (crash_base, crash_size)) { +pr_warn ("crashkernel has wrong address or size\n" ); +return ; + } + +if (crash_base & ((1 <<21 ) -1 )) {
IS_ALIGNED
arch/arm64/mm/init.c static int __init early_initrd(char *p) · line 116
Quoted 13 lines
arch/arm64/mm/init.c +13
-0+pr_warn ("crashkernel base address is not 2MB aligned\n" ); +return ; + } + +memblock_reserve (crash_base, crash_size); + } + +pr_info ("Reserving %lldMB of memory at %lldMB for crashkernel\n" , + crash_size >>20 , crash_base >>20 ); + + crashk_res.start = crash_base; + crashk_res.end = crash_base + crash_size -1 ; +}
#else
static void __init reserve_crashkernel(void) {};
Will
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 11/16] arm64: kdump: reserve memory for crash dump kernel
Quoted 32 lines · Will Deacon
Quoted 31 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
On primary kernel, the memory region used by crash dump kernel must be specified by "crashkernel=" boot parameter.
reserve_crashkernel()will allocate and reserve the region for later use.User space tools will be able to find the region marked as "Crash kernel" in
/proc/iomem.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm64/kernel/setup.c | 7 +++++- arch/arm64/mm/init.c | 54 +++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 60 insertions(+), 1 deletion(-)arch/arm64/kernel/setup.c +0
-1@@ -31,7 +31,6 @@#include <linux/screen_info.h> #include <linux/init.h> #include <linux/kexec.h> -#include <linux/crash_dump.h> #include <linux/root_dev.h> #include <linux/cpu.h> #include <linux/interrupt.h> @@ -221,6 +220,12 @@ static void __init request_standard_resources(void) kernel_data.end <= res->end)request_resource (res, &kernel_data); }
Quoted 8 lines
arch/arm64/kernel/setup.c static void __init request_standard_resources(void) · line 228
Quoted 6 lines
arch/arm64/kernel/setup.c +6
-0+ +#ifdef CONFIG_KEXEC +/* User space tools will find "Crash kernel" region in /proc/iomem. */ +if (crashk_res.end) +insert_resource (&iomem_resource, &crashk_res); +#endif Why can't this bit be moved into
reserved_crashkernel, at the end?
Because
- I want to call reserve_crashkernel() as early as possbile, (so before request_standard_resources()), and
- if inserting "Crash kernel" region is done before request_standard_resources(), request_resource() for "System RAM" region which contains "Crash kernel" will fail and the entry will be lost from /proc/iomem.
Quoted 30 lines
Quoted 30 lines
arch/arm64/kernel/setup.c +0
-0}#ifdef CONFIG_BLK_DEV_INITRD arch/arm64/mm/init.c +12
-0@@ -34,6 +34,7 @@#include <linux/dma-contiguous.h> #include <linux/efi.h> #include <linux/swiotlb.h> +#include <linux/kexec.h> #include <asm/fixmap.h> #include <asm/memory.h> @@ -66,6 +67,55 @@ static int __init early_initrd(char *p)early_param ("initrd" , early_initrd);#endif +#ifdef CONFIG_KEXEC +/* +* reserve_crashkernel() - reserves memory for crash kernel +* +* This function reserves memory area given in "crashkernel=" kernel command +* line parameter. The memory reserved is used by dump capture kernel when +* primary kernel is crashing. +*/ +static void __init reserve_crashkernel (void ) +{ +unsigned long long crash_size =0 , crash_base =0 ;
Quoted 8 lines
arch/arm64/mm/init.c static int __init early_initrd(char *p) · line 86
Quoted 6 lines
arch/arm64/mm/init.c +6
-0+int ret; + + ret =parse_crashkernel (boot_command_line,memblock_phys_mem_size (), + &crash_size, &crash_base); +if (ret) +return ;What's the point in ret?
parse_crashkernel() will returns -EINVAL if
- the boot command line doesn't contain "crashkernel=" parameter, or
- "crashkernel=" parameter has an invalid synstax of value Otherwise returns 0.
So just returning should be OK, but just in case that the syntax is OK, but that a resulting crash_size gets 0, I will add a check like:
if (!crash_size)
return;Quoted 6 lines
arch/arm64/mm/init.c static int __init early_initrd(char *p) · line 89
Quoted 3 lines
arch/arm64/mm/init.c +3
-0+ +if (crash_base ==0 ) { + crash_base =memblock_alloc (crash_size,1 <<21 );What is this magic number? How about
SZ_2Mand a comment mentioning the arm64 boot protocol?
Right. I will do so.
Quoted 8 lines
Quoted 8 lines
arch/arm64/mm/init.c +8
-0+if (crash_base ==0 ) { +pr_warn ("Unable to allocate crashkernel (size:%llx)\n" , + crash_size); +return ; + } + }else { +/* User specifies base address explicitly. */ + if (!memblock_is_region_memory(crash_base, crash_size) ||
Quoted 8 lines
arch/arm64/mm/init.c static int __init early_initrd(char *p) · line 103
Quoted 6 lines
arch/arm64/mm/init.c +6
-0+ memblock_is_region_reserved(crash_base, crash_size)) { +pr_warn ("crashkernel has wrong address or size\n" ); +return ; + } + +if (crash_base & ((1 <<21 ) -1 )) {
IS_ALIGNED
OK.
Quoted 16 lines
arch/arm64/mm/init.c static int __init early_initrd(char *p) · line 116
Quoted 13 lines
arch/arm64/mm/init.c +13
-0+pr_warn ("crashkernel base address is not 2MB aligned\n" ); +return ; + } + +memblock_reserve (crash_base, crash_size); + } + +pr_info ("Reserving %lldMB of memory at %lldMB for crashkernel\n" , + crash_size >>20 , crash_base >>20 ); + + crashk_res.start = crash_base; + crashk_res.end = crash_base + crash_size -1 ; +}#else
Quoted 1 line
static void
__initreserve_crashkernel(void){};
OK.
Thanks,
-Takahiro AKASHI
Quoted 2 lines
Will
Re: [PATCH v12 11/16] arm64: kdump: reserve memory for crash dump kernel
Quoted 6 lines · AKASHI Takahiro
Quoted 3 lines · Will Deacon
#else
static void
__initreserve_crashkernel(void){};OK.
Then may be you do not need to protect calling of reserve_crashkernel() with CONFIG_KEXEC_CORE as well.
~Pratyush
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 12/16 arm64: kdump: implement machine_crash_shutdown()
In reply to Geoff LevandFrom: AKASHI Takahiro <[email protected]>
kdump calls machine_crash_shutdown() to shut down non-boot cpus and save registers' status in per-cpu ELF notes before starting the crash dump kernel. See kernel_kexec().
ipi_cpu_stop() is a bit modified and used to support this behavior.
--- arch/arm64/include/asm/kexec.h | 34 +++++++++++++++++++++++++++++++++- arch/arm64/kernel/machine_kexec.c | 31 +++++++++++++++++++++++++++++-- arch/arm64/kernel/smp.c | 16 ++++++++++++++-- 3 files changed, 76 insertions(+), 5 deletions(-)
arch/arm64/include/asm/kexec.h +33 -1
@@ -30,6 +30,8 @@#if !defined(__ASSEMBLY__) +extern bool in_crash_kexec; +/** * crash_setup_regs() - save registers for the panic kernel *
@@ -40,7 +42,37 @@static inline void crash_setup_regs (struct pt_regs *newregs,struct pt_regs *oldregs) { -/* Empty routine needed to avoid build errors. */ +if (oldregs) { +memcpy (newregs, oldregs,sizeof (*newregs)); + }else { + __asm__ __volatile__ ( +"stp x0, x1, [%3, #16 * 0]\n" +"stp x2, x3, [%3, #16 * 1]\n" +"stp x4, x5, [%3, #16 * 2]\n" +"stp x6, x7, [%3, #16 * 3]\n" +"stp x8, x9, [%3, #16 * 4]\n" +"stp x10, x11, [%3, #16 * 5]\n" +"stp x12, x13, [%3, #16 * 6]\n" +"stp x14, x15, [%3, #16 * 7]\n" +"stp x16, x17, [%3, #16 * 8]\n" +"stp x18, x19, [%3, #16 * 9]\n" +"stp x20, x21, [%3, #16 * 10]\n" +"stp x22, x23, [%3, #16 * 11]\n" +"stp x24, x25, [%3, #16 * 12]\n" +"stp x26, x27, [%3, #16 * 13]\n" +"stp x28, x29, [%3, #16 * 14]\n" +"str x30, [%3, #16 * 15]\n" +"mov %0, sp\n" +"adr %1, 1f\n" +"mrs %2, spsr_el1\n" +"1:" + :"=r" (newregs->sp), +"=r" (newregs->pc), +"=r" (newregs->pstate) + :"r" (&newregs->regs) + :"memory" + ); + } }#endif /* !defined(__ASSEMBLY__) */
arch/arm64/kernel/machine_kexec.c +29 -2
@@ -9,6 +9,7 @@* published by the Free Software Foundation. */ +#include <linux/kernel.h> #include <linux/kexec.h> #include <linux/of_fdt.h> #include <linux/slab.h>
@@ -23,6 +24,7 @@extern const unsigned char arm64_relocate_new_kernel[];extern const unsigned long arm64_relocate_new_kernel_size; +bool in_crash_kexec;static unsigned long kimage_start;/**
@@ -203,13 +205,38 @@ void machine_kexec(struct kimage *kimage)*/ cpu_soft_restart (virt_to_phys (cpu_reset), -is_hyp_mode_available (), + in_crash_kexec ?0 :is_hyp_mode_available (), reboot_code_buffer_phys, kimage->head, kimage_start);BUG ();/* Should never get here. */ } +/** +* machine_crash_shutdown - shutdown non-boot cpus and save registers +*/ void machine_crash_shutdown (struct pt_regs *regs) { -/* Empty routine needed to avoid build errors. */ +struct pt_regs dummy_regs; +int cpu; + +local_irq_disable (); + + in_crash_kexec = true; + +/* +* clear and initialize the per-cpu info. This is necessary +* because, otherwise, slots for offline cpus would never be +* filled up. See smp_send_stop(). +*/ +memset (&dummy_regs,0 ,sizeof (dummy_regs)); +for_each_possible_cpu (cpu) +crash_save_cpu (&dummy_regs, cpu); + +/* shutdown non-boot cpus */ +smp_send_stop (); + +/* for boot cpu */ +crash_save_cpu (regs,smp_processor_id ()); + +pr_info ("Starting crashdump kernel...\n" ); }
arch/arm64/kernel/smp.c +14 -2
@@ -37,6 +37,7 @@#include <linux/completion.h> #include <linux/of.h> #include <linux/irq_work.h> +#include <linux/kexec.h> #include <asm/alternative.h> #include <asm/atomic.h>
@@ -54,6 +55,8 @@#include <asm/ptrace.h> #include <asm/virt.h> +#include "cpu-reset.h" +#define CREATE_TRACE_POINTS #include <trace/events/ipi.h>
@@ -683,8 +686,12 @@ static DEFINE_RAW_SPINLOCK(stop_lock);/* * ipi_cpu_stop - handle IPI from smp_send_stop() */ -static void ipi_cpu_stop (unsigned int cpu) +static void ipi_cpu_stop (unsigned int cpu,struct pt_regs *regs) { +#ifdef CONFIG_KEXEC +/* printing messages may slow down the shutdown. */ + if (!in_crash_kexec) +#endif if (system_state ==SYSTEM_BOOTING || system_state ==SYSTEM_RUNNING ) {raw_spin_lock (&stop_lock);
@@ -697,6 +704,11 @@ static void ipi_cpu_stop(unsigned int cpu)local_irq_disable (); +#ifdef CONFIG_KEXEC +if (in_crash_kexec) +crash_save_cpu (regs, cpu); +#endif /* CONFIG_KEXEC */ +while (1 )cpu_relax (); }
@@ -727,7 +739,7 @@ void handle_IPI(int ipinr, struct pt_regs *regs)case IPI_CPU_STOP :irq_enter (); -ipi_cpu_stop (cpu); +ipi_cpu_stop (cpu, regs);irq_exit ();break ;
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 12/16] arm64: kdump: implement machine_crash_shutdown()
Quoted 53 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
kdump calls
machine_crash_shutdown()to shut down non-boot cpus and save registers' status in per-cpu ELF notes before starting the crash dump kernel. Seekernel_kexec().
ipi_cpu_stop()is a bit modified and used to support this behavior.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm64/include/asm/kexec.h | 34 +++++++++++++++++++++++++++++++++- arch/arm64/kernel/machine_kexec.c | 31 +++++++++++++++++++++++++++++-- arch/arm64/kernel/smp.c | 16 ++++++++++++++-- 3 files changed, 76 insertions(+), 5 deletions(-)arch/arm64/include/asm/kexec.h +22
-1@@ -30,6 +30,8 @@#if !defined(__ASSEMBLY__) +extern bool in_crash_kexec; +/** * crash_setup_regs() - save registers for the panic kernel * @@ -40,7 +42,37 @@static inline void crash_setup_regs (struct pt_regs *newregs,struct pt_regs *oldregs) { -/* Empty routine needed to avoid build errors. */ +if (oldregs) { +memcpy (newregs, oldregs,sizeof (*newregs)); + }else { + __asm__ __volatile__ ( +"stp x0, x1, [%3, #16 * 0]\n" +"stp x2, x3, [%3, #16 * 1]\n" +"stp x4, x5, [%3, #16 * 2]\n" +"stp x6, x7, [%3, #16 * 3]\n" +"stp x8, x9, [%3, #16 * 4]\n" +"stp x10, x11, [%3, #16 * 5]\n" +"stp x12, x13, [%3, #16 * 6]\n" +"stp x14, x15, [%3, #16 * 7]\n" +"stp x16, x17, [%3, #16 * 8]\n" +"stp x18, x19, [%3, #16 * 9]\n" +"stp x20, x21, [%3, #16 * 10]\n" +"stp x22, x23, [%3, #16 * 11]\n" +"stp x24, x25, [%3, #16 * 12]\n" +"stp x26, x27, [%3, #16 * 13]\n" +"stp x28, x29, [%3, #16 * 14]\n" + "str x30, [%3, #16 * 15]\n"
arch/arm64/include/asm/kexec.h line 74
Quoted 10 lines
arch/arm64/include/asm/kexec.h +10
-0+ "mov %0, sp\n" + "adr %1, 1f\n" + "mrs %2, spsr_el1\n" + "1:" + : "=r" (newregs->sp), + "=r" (newregs->pc), + "=r" (newregs->pstate) + : "r" (&newregs->regs) + : "memory" + );
I wonder how useful this thing is, given that it starts by corrupting whatever register is holding newregs->regs. Maybe this is not supposed to be accurate anyway...
Quoted 90 lines
arch/arm64/include/asm/kexec.h +1
-0+ } }#endif /* !defined(__ASSEMBLY__) */ arch/arm64/kernel/machine_kexec.c +29
-2@@ -9,6 +9,7 @@* published by the Free Software Foundation. */ +#include <linux/kernel.h> #include <linux/kexec.h> #include <linux/of_fdt.h> #include <linux/slab.h> @@ -23,6 +24,7 @@extern const unsigned char arm64_relocate_new_kernel[];extern const unsigned long arm64_relocate_new_kernel_size; +bool in_crash_kexec;static unsigned long kimage_start;/** @@ -203,13 +205,38 @@ void machine_kexec(struct kimage *kimage)*/ cpu_soft_restart (virt_to_phys (cpu_reset), -is_hyp_mode_available (), + in_crash_kexec ?0 :is_hyp_mode_available (), reboot_code_buffer_phys, kimage->head, kimage_start);BUG ();/* Should never get here. */ } +/** +* machine_crash_shutdown - shutdown non-boot cpus and save registers +*/ void machine_crash_shutdown (struct pt_regs *regs) { -/* Empty routine needed to avoid build errors. */ +struct pt_regs dummy_regs; +int cpu; + +local_irq_disable (); + + in_crash_kexec = true; + +/* +* clear and initialize the per-cpu info. This is necessary +* because, otherwise, slots for offline cpus would never be +* filled up. See smp_send_stop(). +*/ +memset (&dummy_regs,0 ,sizeof (dummy_regs)); +for_each_possible_cpu (cpu) +crash_save_cpu (&dummy_regs, cpu); + +/* shutdown non-boot cpus */ +smp_send_stop (); + +/* for boot cpu */ +crash_save_cpu (regs,smp_processor_id ()); + +pr_info ("Starting crashdump kernel...\n" ); }arch/arm64/kernel/smp.c +3
-0@@ -37,6 +37,7 @@#include <linux/completion.h> #include <linux/of.h> #include <linux/irq_work.h> +#include <linux/kexec.h> #include <asm/alternative.h> #include <asm/atomic.h> @@ -54,6 +55,8 @@#include <asm/ptrace.h> #include <asm/virt.h> +#include "cpu-reset.h" +#define CREATE_TRACE_POINTS #include <trace/events/ipi.h> @@ -683,8 +686,12 @@ static DEFINE_RAW_SPINLOCK(stop_lock);/* * ipi_cpu_stop - handle IPI from smp_send_stop() */
arch/arm64/kernel/smp.c static DEFINE_RAW_SPINLOCK(stop_lock); · line 697
Quoted 10 lines
arch/arm64/kernel/smp.c +5
-1-static void ipi_cpu_stop (unsigned int cpu) +static void ipi_cpu_stop (unsigned int cpu,struct pt_regs *regs) { +#ifdef CONFIG_KEXEC +/* printing messages may slow down the shutdown. */ + if (!in_crash_kexec) +#endif if (system_state ==SYSTEM_BOOTING || system_state ==SYSTEM_RUNNING ) {raw_spin_lock (&stop_lock);
Irrespective of how useful this change is, how about having a predicate instead? Something like:
static inline bool is_in_crash_kexec(void)
{
#ifdef CONFIG_KEXEC return in_crash_kexec; #else return false;
#endif
}
located in machine_kexec.c (making the in_crash_kernel static), and then
if (!is_in_crash_kexec() && (systen_state == ... || ...) {It would certainly look better.
arch/arm64/kernel/smp.c static void ipi_cpu_stop(unsigned int cpu) · line 711
Quoted 9 lines
arch/arm64/kernel/smp.c +5
-0@@ -697,6 +704,11 @@ static void ipi_cpu_stop(unsigned int cpu)local_irq_disable (); +#ifdef CONFIG_KEXEC +if (in_crash_kexec) +crash_save_cpu (regs, cpu); +#endif /* CONFIG_KEXEC */ +
Same here.
arch/arm64/kernel/smp.c void handle_IPI(int ipinr, struct pt_regs *regs) · line 745
Quoted 13 lines
arch/arm64/kernel/smp.c +1
-1while (1 )cpu_relax (); }@@ -727,7 +739,7 @@ void handle_IPI(int ipinr, struct pt_regs *regs)case IPI_CPU_STOP :irq_enter (); -ipi_cpu_stop (cpu); +ipi_cpu_stop (cpu, regs);irq_exit ();break ;
Thanks,
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 12/16] arm64: kdump: implement machine_crash_shutdown()
Marc,
I was back from my vacation.
Quoted 60 lines · Marc Zyngier
Quoted 59 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
kdump calls
machine_crash_shutdown()to shut down non-boot cpus and save registers' status in per-cpu ELF notes before starting the crash dump kernel. Seekernel_kexec().
ipi_cpu_stop()is a bit modified and used to support this behavior.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm64/include/asm/kexec.h | 34 +++++++++++++++++++++++++++++++++- arch/arm64/kernel/machine_kexec.c | 31 +++++++++++++++++++++++++++++-- arch/arm64/kernel/smp.c | 16 ++++++++++++++-- 3 files changed, 76 insertions(+), 5 deletions(-)arch/arm64/include/asm/kexec.h +28
-1@@ -30,6 +30,8 @@#if !defined(__ASSEMBLY__) +extern bool in_crash_kexec; +/** * crash_setup_regs() - save registers for the panic kernel * @@ -40,7 +42,37 @@static inline void crash_setup_regs (struct pt_regs *newregs,struct pt_regs *oldregs) { -/* Empty routine needed to avoid build errors. */ +if (oldregs) { +memcpy (newregs, oldregs,sizeof (*newregs)); + }else { + __asm__ __volatile__ ( + "stp x0, x1, [%3, #16 * 0]\n" + "stp x2, x3, [%3, #16 * 1]\n" + "stp x4, x5, [%3, #16 * 2]\n" + "stp x6, x7, [%3, #16 * 3]\n" + "stp x8, x9, [%3, #16 * 4]\n" + "stp x10, x11, [%3, #16 * 5]\n" + "stp x12, x13, [%3, #16 * 6]\n" + "stp x14, x15, [%3, #16 * 7]\n" + "stp x16, x17, [%3, #16 * 8]\n" + "stp x18, x19, [%3, #16 * 9]\n" + "stp x20, x21, [%3, #16 * 10]\n" + "stp x22, x23, [%3, #16 * 11]\n" + "stp x24, x25, [%3, #16 * 12]\n" + "stp x26, x27, [%3, #16 * 13]\n" + "stp x28, x29, [%3, #16 * 14]\n" + "str x30, [%3, #16 * 15]\n" + "mov %0, sp\n" + "adr %1, 1f\n" + "mrs %2, spsr_el1\n" + "1:" + : "=r" (newregs->sp), + "=r" (newregs->pc),
Quoted 8 lines
arch/arm64/include/asm/kexec.h line 74
Quoted 4 lines
arch/arm64/include/asm/kexec.h +4
-0+ "=r" (newregs->pstate) + : "r" (&newregs->regs) + : "memory" + );I wonder how useful this thing is, given that it starts by corrupting whatever register is holding
newregs->regs. Maybe this is not supposed to be accurate anyway...
I'm not quite sure about what part of my code you're mentioning here, but crash_setup_regs() is solely called by crash_kexec(), and panic() is the only caller of crash_kexec() with NULL argument which, in turn, is used as 'oldregs' in crash_setup_regs().
Given this fact, I think that the values saved in newregs as indicated above will be the best estimate of current cpu contexts.
The other caller of crash_kexec() is die() in traps.c, but here we call it with explicit cpu contexts at exception.
Quoted 105 lines
Quoted 91 lines
arch/arm64/include/asm/kexec.h +1
-0+ } }#endif /* !defined(__ASSEMBLY__) */ arch/arm64/kernel/machine_kexec.c +29
-2@@ -9,6 +9,7 @@* published by the Free Software Foundation. */ +#include <linux/kernel.h> #include <linux/kexec.h> #include <linux/of_fdt.h> #include <linux/slab.h> @@ -23,6 +24,7 @@extern const unsigned char arm64_relocate_new_kernel[];extern const unsigned long arm64_relocate_new_kernel_size; +bool in_crash_kexec;static unsigned long kimage_start;/** @@ -203,13 +205,38 @@ void machine_kexec(struct kimage *kimage)*/ cpu_soft_restart (virt_to_phys (cpu_reset), -is_hyp_mode_available (), + in_crash_kexec ?0 :is_hyp_mode_available (), reboot_code_buffer_phys, kimage->head, kimage_start);BUG ();/* Should never get here. */ } +/** +* machine_crash_shutdown - shutdown non-boot cpus and save registers +*/ void machine_crash_shutdown (struct pt_regs *regs) { -/* Empty routine needed to avoid build errors. */ +struct pt_regs dummy_regs; +int cpu; + +local_irq_disable (); + + in_crash_kexec = true; + +/* +* clear and initialize the per-cpu info. This is necessary +* because, otherwise, slots for offline cpus would never be +* filled up. See smp_send_stop(). +*/ +memset (&dummy_regs,0 ,sizeof (dummy_regs)); +for_each_possible_cpu (cpu) +crash_save_cpu (&dummy_regs, cpu); + +/* shutdown non-boot cpus */ +smp_send_stop (); + +/* for boot cpu */ +crash_save_cpu (regs,smp_processor_id ()); + +pr_info ("Starting crashdump kernel...\n" ); }arch/arm64/kernel/smp.c +3
-0@@ -37,6 +37,7 @@#include <linux/completion.h> #include <linux/of.h> #include <linux/irq_work.h> +#include <linux/kexec.h> #include <asm/alternative.h> #include <asm/atomic.h> @@ -54,6 +55,8 @@#include <asm/ptrace.h> #include <asm/virt.h> +#include "cpu-reset.h" +#define CREATE_TRACE_POINTS #include <trace/events/ipi.h> @@ -683,8 +686,12 @@ static DEFINE_RAW_SPINLOCK(stop_lock);/* * ipi_cpu_stop - handle IPI from smp_send_stop() */ arch/arm64/kernel/smp.c static DEFINE_RAW_SPINLOCK(stop_lock); · line 697
Quoted 10 lines
arch/arm64/kernel/smp.c +5
-1-static void ipi_cpu_stop (unsigned int cpu) +static void ipi_cpu_stop (unsigned int cpu,struct pt_regs *regs) { +#ifdef CONFIG_KEXEC +/* printing messages may slow down the shutdown. */ + if (!in_crash_kexec) +#endif if (system_state ==SYSTEM_BOOTING || system_state ==SYSTEM_RUNNING ) {raw_spin_lock (&stop_lock);Irrespective of how useful this change is, how about having a predicate instead? Something like:
Quoted 8 lines
static inline bool
is_in_crash_kexec(void)
{#ifdef CONFIG_KEXEC return in_crash_kexec; #else return false;#endif
}
OK, I will take your idea.
Quoted 1 line
located in
machine_kexec.c(making thein_crash_kernelstatic), and then
but cannot make in_crash_kernel static because it is also used in both smp.c and machine_kexec.c.
Quoted 3 lines
if (!is_in_crash_kexec() && (systen_state == ... || ...) {It would certainly look better.
Thanks,
-Takahiro AKASHI
Quoted 30 lines
arch/arm64/kernel/smp.c static void ipi_cpu_stop(unsigned int cpu) · line 711
Quoted 9 lines
arch/arm64/kernel/smp.c +5
-0@@ -697,6 +704,11 @@ static void ipi_cpu_stop(unsigned int cpu)local_irq_disable (); +#ifdef CONFIG_KEXEC +if (in_crash_kexec) +crash_save_cpu (regs, cpu); +#endif /* CONFIG_KEXEC */ +Same here.
arch/arm64/kernel/smp.c void handle_IPI(int ipinr, struct pt_regs *regs) · line 745
Quoted 13 lines
arch/arm64/kernel/smp.c +1
-1while (1 )cpu_relax (); }@@ -727,7 +739,7 @@ void handle_IPI(int ipinr, struct pt_regs *regs)case IPI_CPU_STOP :irq_enter (); -ipi_cpu_stop (cpu); +ipi_cpu_stop (cpu, regs);irq_exit ();break ;Thanks,
M.
Re: [PATCH v12 12/16] arm64: kdump: implement machine_crash_shutdown()
Quoted 74 lines · AKASHI Takahiro
Marc,
I was back from my vacation.
Quoted 60 lines · Marc Zyngier
Quoted 59 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
kdump calls
machine_crash_shutdown()to shut down non-boot cpus and save registers' status in per-cpu ELF notes before starting the crash dump kernel. Seekernel_kexec().
ipi_cpu_stop()is a bit modified and used to support this behavior.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm64/include/asm/kexec.h | 34 +++++++++++++++++++++++++++++++++- arch/arm64/kernel/machine_kexec.c | 31 +++++++++++++++++++++++++++++-- arch/arm64/kernel/smp.c | 16 ++++++++++++++-- 3 files changed, 76 insertions(+), 5 deletions(-)arch/arm64/include/asm/kexec.h +28
-1@@ -30,6 +30,8 @@#if !defined(__ASSEMBLY__) +extern bool in_crash_kexec; +/** * crash_setup_regs() - save registers for the panic kernel * @@ -40,7 +42,37 @@static inline void crash_setup_regs (struct pt_regs *newregs,struct pt_regs *oldregs) { -/* Empty routine needed to avoid build errors. */ +if (oldregs) { +memcpy (newregs, oldregs,sizeof (*newregs)); + }else { + __asm__ __volatile__ ( + "stp x0, x1, [%3, #16 * 0]\n" + "stp x2, x3, [%3, #16 * 1]\n" + "stp x4, x5, [%3, #16 * 2]\n" + "stp x6, x7, [%3, #16 * 3]\n" + "stp x8, x9, [%3, #16 * 4]\n" + "stp x10, x11, [%3, #16 * 5]\n" + "stp x12, x13, [%3, #16 * 6]\n" + "stp x14, x15, [%3, #16 * 7]\n" + "stp x16, x17, [%3, #16 * 8]\n" + "stp x18, x19, [%3, #16 * 9]\n" + "stp x20, x21, [%3, #16 * 10]\n" + "stp x22, x23, [%3, #16 * 11]\n" + "stp x24, x25, [%3, #16 * 12]\n" + "stp x26, x27, [%3, #16 * 13]\n" + "stp x28, x29, [%3, #16 * 14]\n" + "str x30, [%3, #16 * 15]\n" + "mov %0, sp\n" + "adr %1, 1f\n" + "mrs %2, spsr_el1\n" + "1:" + : "=r" (newregs->sp), + "=r" (newregs->pc),Quoted 8 lines
arch/arm64/include/asm/kexec.h line 74
Quoted 4 lines
arch/arm64/include/asm/kexec.h +4
-0+ "=r" (newregs->pstate) + : "r" (&newregs->regs) + : "memory" + );I wonder how useful this thing is, given that it starts by corrupting whatever register is holding
newregs->regs. Maybe this is not supposed to be accurate anyway...
Quoted 4 lines
I'm not quite sure about what part of my code you're mentioning here, but
crash_setup_regs()is solely called bycrash_kexec(), andpanic()is the only caller ofcrash_kexec()with NULL argument which, in turn, is used as 'oldregs' incrash_setup_regs().
You have this assembly sequence:
stp x0, x1, [%3, #16 * 0]
[...]
where %3 itself is one of the x[0..30] registers. So you are saving things that have already been corrupted by the saving procedure. Not sure how useful that is, but as I said, maybe it is not supposed to be completely accurate.
Quoted 120 lines
Given this fact, I think that the values saved in newregs as indicated above will be the best estimate of current cpu contexts.
The other caller of
crash_kexec()isdie()intraps.c, but here we call it with explicit cpu contexts at exception.Quoted 105 lines
Quoted 91 lines
arch/arm64/include/asm/kexec.h +1
-0+ } }#endif /* !defined(__ASSEMBLY__) */ arch/arm64/kernel/machine_kexec.c +29
-2@@ -9,6 +9,7 @@* published by the Free Software Foundation. */ +#include <linux/kernel.h> #include <linux/kexec.h> #include <linux/of_fdt.h> #include <linux/slab.h> @@ -23,6 +24,7 @@extern const unsigned char arm64_relocate_new_kernel[];extern const unsigned long arm64_relocate_new_kernel_size; +bool in_crash_kexec;static unsigned long kimage_start;/** @@ -203,13 +205,38 @@ void machine_kexec(struct kimage *kimage)*/ cpu_soft_restart (virt_to_phys (cpu_reset), -is_hyp_mode_available (), + in_crash_kexec ?0 :is_hyp_mode_available (), reboot_code_buffer_phys, kimage->head, kimage_start);BUG ();/* Should never get here. */ } +/** +* machine_crash_shutdown - shutdown non-boot cpus and save registers +*/ void machine_crash_shutdown (struct pt_regs *regs) { -/* Empty routine needed to avoid build errors. */ +struct pt_regs dummy_regs; +int cpu; + +local_irq_disable (); + + in_crash_kexec = true; + +/* +* clear and initialize the per-cpu info. This is necessary +* because, otherwise, slots for offline cpus would never be +* filled up. See smp_send_stop(). +*/ +memset (&dummy_regs,0 ,sizeof (dummy_regs)); +for_each_possible_cpu (cpu) +crash_save_cpu (&dummy_regs, cpu); + +/* shutdown non-boot cpus */ +smp_send_stop (); + +/* for boot cpu */ +crash_save_cpu (regs,smp_processor_id ()); + +pr_info ("Starting crashdump kernel...\n" ); }arch/arm64/kernel/smp.c +3
-0@@ -37,6 +37,7 @@#include <linux/completion.h> #include <linux/of.h> #include <linux/irq_work.h> +#include <linux/kexec.h> #include <asm/alternative.h> #include <asm/atomic.h> @@ -54,6 +55,8 @@#include <asm/ptrace.h> #include <asm/virt.h> +#include "cpu-reset.h" +#define CREATE_TRACE_POINTS #include <trace/events/ipi.h> @@ -683,8 +686,12 @@ static DEFINE_RAW_SPINLOCK(stop_lock);/* * ipi_cpu_stop - handle IPI from smp_send_stop() */ arch/arm64/kernel/smp.c static DEFINE_RAW_SPINLOCK(stop_lock); · line 697
Quoted 10 lines
arch/arm64/kernel/smp.c +5
-1-static void ipi_cpu_stop (unsigned int cpu) +static void ipi_cpu_stop (unsigned int cpu,struct pt_regs *regs) { +#ifdef CONFIG_KEXEC +/* printing messages may slow down the shutdown. */ + if (!in_crash_kexec) +#endif if (system_state ==SYSTEM_BOOTING || system_state ==SYSTEM_RUNNING ) {raw_spin_lock (&stop_lock);Irrespective of how useful this change is, how about having a predicate instead? Something like:
Quoted 8 lines
static inline bool
is_in_crash_kexec(void)
{#ifdef CONFIG_KEXEC return in_crash_kexec; #else return false;#endif
}
Quoted 6 lines
OK, I will take your idea.
Quoted 1 line
located in
machine_kexec.c(making thein_crash_kernelstatic), and thenbut cannot make
in_crash_kernelstatic because it is also used in bothsmp.candmachine_kexec.c.
smp.c only reads from in_crash_kernel (at least from what I can see in this patch), so it should be able to use the accessor.
Thanks,
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 12/16] arm64: kdump: implement machine_crash_shutdown()
Quoted 80 lines · Marc Zyngier
Quoted 74 lines · AKASHI Takahiro
Marc,
I was back from my vacation.
Quoted 60 lines · Marc Zyngier
Quoted 59 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
kdump calls
machine_crash_shutdown()to shut down non-boot cpus and save registers' status in per-cpu ELF notes before starting the crash dump kernel. Seekernel_kexec().
ipi_cpu_stop()is a bit modified and used to support this behavior.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm64/include/asm/kexec.h | 34 +++++++++++++++++++++++++++++++++- arch/arm64/kernel/machine_kexec.c | 31 +++++++++++++++++++++++++++++-- arch/arm64/kernel/smp.c | 16 ++++++++++++++-- 3 files changed, 76 insertions(+), 5 deletions(-)arch/arm64/include/asm/kexec.h +28
-1@@ -30,6 +30,8 @@#if !defined(__ASSEMBLY__) +extern bool in_crash_kexec; +/** * crash_setup_regs() - save registers for the panic kernel * @@ -40,7 +42,37 @@static inline void crash_setup_regs (struct pt_regs *newregs,struct pt_regs *oldregs) { -/* Empty routine needed to avoid build errors. */ +if (oldregs) { +memcpy (newregs, oldregs,sizeof (*newregs)); + }else { + __asm__ __volatile__ ( + "stp x0, x1, [%3, #16 * 0]\n" + "stp x2, x3, [%3, #16 * 1]\n" + "stp x4, x5, [%3, #16 * 2]\n" + "stp x6, x7, [%3, #16 * 3]\n" + "stp x8, x9, [%3, #16 * 4]\n" + "stp x10, x11, [%3, #16 * 5]\n" + "stp x12, x13, [%3, #16 * 6]\n" + "stp x14, x15, [%3, #16 * 7]\n" + "stp x16, x17, [%3, #16 * 8]\n" + "stp x18, x19, [%3, #16 * 9]\n" + "stp x20, x21, [%3, #16 * 10]\n" + "stp x22, x23, [%3, #16 * 11]\n" + "stp x24, x25, [%3, #16 * 12]\n" + "stp x26, x27, [%3, #16 * 13]\n" + "stp x28, x29, [%3, #16 * 14]\n" + "str x30, [%3, #16 * 15]\n" + "mov %0, sp\n" + "adr %1, 1f\n" + "mrs %2, spsr_el1\n" + "1:" + : "=r" (newregs->sp), + "=r" (newregs->pc),Quoted 8 lines
arch/arm64/include/asm/kexec.h line 74
Quoted 4 lines
arch/arm64/include/asm/kexec.h +4
-0+ "=r" (newregs->pstate) + : "r" (&newregs->regs) + : "memory" + );I wonder how useful this thing is, given that it starts by corrupting whatever register is holding
newregs->regs. Maybe this is not supposed to be accurate anyway...Quoted 4 lines
I'm not quite sure about what part of my code you're mentioning here, but
crash_setup_regs()is solely called bycrash_kexec(), andpanic()is the only caller ofcrash_kexec()with NULL argument which, in turn, is used as 'oldregs' incrash_setup_regs().
Quoted 6 lines
You have this assembly sequence:
stp x0, x1, [%3, #16 * 0]
[...]where %3 itself is one of the x[0..30] registers.
Not only %3, but also
Quoted 4 lines
So you are saving
things that have already been corrupted by the saving procedure. Not sure how useful that is, but as I said, maybe it is not supposed to be completely accurate.
x0, x1 ... are the current values in panic(), and not the exact cpu contexts at the place we are really interested in.
We have no way here in panic() to know them, but sp and pc would still be useful for back-tracing in later investigation of dump file.
Please note that the same problem exists on arm (and x86) implementation.
Quoted 122 lines
Quoted 122 lines
Given this fact, I think that the values saved in newregs as indicated above will be the best estimate of current cpu contexts.
The other caller of
crash_kexec()isdie()intraps.c, but here we call it with explicit cpu contexts at exception.Quoted 105 lines
Quoted 91 lines
arch/arm64/include/asm/kexec.h +1
-0+ } }#endif /* !defined(__ASSEMBLY__) */ arch/arm64/kernel/machine_kexec.c +29
-2@@ -9,6 +9,7 @@* published by the Free Software Foundation. */ +#include <linux/kernel.h> #include <linux/kexec.h> #include <linux/of_fdt.h> #include <linux/slab.h> @@ -23,6 +24,7 @@extern const unsigned char arm64_relocate_new_kernel[];extern const unsigned long arm64_relocate_new_kernel_size; +bool in_crash_kexec;static unsigned long kimage_start;/** @@ -203,13 +205,38 @@ void machine_kexec(struct kimage *kimage)*/ cpu_soft_restart (virt_to_phys (cpu_reset), -is_hyp_mode_available (), + in_crash_kexec ?0 :is_hyp_mode_available (), reboot_code_buffer_phys, kimage->head, kimage_start);BUG ();/* Should never get here. */ } +/** +* machine_crash_shutdown - shutdown non-boot cpus and save registers +*/ void machine_crash_shutdown (struct pt_regs *regs) { -/* Empty routine needed to avoid build errors. */ +struct pt_regs dummy_regs; +int cpu; + +local_irq_disable (); + + in_crash_kexec = true; + +/* +* clear and initialize the per-cpu info. This is necessary +* because, otherwise, slots for offline cpus would never be +* filled up. See smp_send_stop(). +*/ +memset (&dummy_regs,0 ,sizeof (dummy_regs)); +for_each_possible_cpu (cpu) +crash_save_cpu (&dummy_regs, cpu); + +/* shutdown non-boot cpus */ +smp_send_stop (); + +/* for boot cpu */ +crash_save_cpu (regs,smp_processor_id ()); + +pr_info ("Starting crashdump kernel...\n" ); }arch/arm64/kernel/smp.c +3
-0@@ -37,6 +37,7 @@#include <linux/completion.h> #include <linux/of.h> #include <linux/irq_work.h> +#include <linux/kexec.h> #include <asm/alternative.h> #include <asm/atomic.h> @@ -54,6 +55,8 @@#include <asm/ptrace.h> #include <asm/virt.h> +#include "cpu-reset.h" +#define CREATE_TRACE_POINTS #include <trace/events/ipi.h> @@ -683,8 +686,12 @@ static DEFINE_RAW_SPINLOCK(stop_lock);/* * ipi_cpu_stop - handle IPI from smp_send_stop() */ arch/arm64/kernel/smp.c static DEFINE_RAW_SPINLOCK(stop_lock); · line 697
Quoted 10 lines
arch/arm64/kernel/smp.c +5
-1-static void ipi_cpu_stop (unsigned int cpu) +static void ipi_cpu_stop (unsigned int cpu,struct pt_regs *regs) { +#ifdef CONFIG_KEXEC +/* printing messages may slow down the shutdown. */ + if (!in_crash_kexec) +#endif if (system_state ==SYSTEM_BOOTING || system_state ==SYSTEM_RUNNING ) {raw_spin_lock (&stop_lock);Irrespective of how useful this change is, how about having a predicate instead? Something like:
Quoted 8 lines
static inline bool
is_in_crash_kexec(void)
{#ifdef CONFIG_KEXEC return in_crash_kexec; #else return false;#endif
}OK, I will take your idea.
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
located in
machine_kexec.c(making thein_crash_kernelstatic), and thenbut cannot make
in_crash_kernelstatic because it is also used in bothsmp.candmachine_kexec.c.
smp.conly reads fromin_crash_kernel(at least from what I can see in this patch), so it should be able to use the accessor.
Only if we define the accessor as a real function, not an inline function in a header :)
-Takahiro AKASHI
Quoted 4 lines
Thanks,
M.
Re: [PATCH v12 12/16] arm64: kdump: implement machine_crash_shutdown()
Quoted 95 lines · AKASHI Takahiro
Quoted 80 lines · Marc Zyngier
Quoted 74 lines · AKASHI Takahiro
Marc,
I was back from my vacation.
Quoted 60 lines · Marc Zyngier
Quoted 59 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
kdump calls
machine_crash_shutdown()to shut down non-boot cpus and save registers' status in per-cpu ELF notes before starting the crash dump kernel. Seekernel_kexec().
ipi_cpu_stop()is a bit modified and used to support this behavior.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm64/include/asm/kexec.h | 34 +++++++++++++++++++++++++++++++++- arch/arm64/kernel/machine_kexec.c | 31 +++++++++++++++++++++++++++++-- arch/arm64/kernel/smp.c | 16 ++++++++++++++-- 3 files changed, 76 insertions(+), 5 deletions(-)arch/arm64/include/asm/kexec.h +28
-1@@ -30,6 +30,8 @@#if !defined(__ASSEMBLY__) +extern bool in_crash_kexec; +/** * crash_setup_regs() - save registers for the panic kernel * @@ -40,7 +42,37 @@static inline void crash_setup_regs (struct pt_regs *newregs,struct pt_regs *oldregs) { -/* Empty routine needed to avoid build errors. */ +if (oldregs) { +memcpy (newregs, oldregs,sizeof (*newregs)); + }else { + __asm__ __volatile__ ( + "stp x0, x1, [%3, #16 * 0]\n" + "stp x2, x3, [%3, #16 * 1]\n" + "stp x4, x5, [%3, #16 * 2]\n" + "stp x6, x7, [%3, #16 * 3]\n" + "stp x8, x9, [%3, #16 * 4]\n" + "stp x10, x11, [%3, #16 * 5]\n" + "stp x12, x13, [%3, #16 * 6]\n" + "stp x14, x15, [%3, #16 * 7]\n" + "stp x16, x17, [%3, #16 * 8]\n" + "stp x18, x19, [%3, #16 * 9]\n" + "stp x20, x21, [%3, #16 * 10]\n" + "stp x22, x23, [%3, #16 * 11]\n" + "stp x24, x25, [%3, #16 * 12]\n" + "stp x26, x27, [%3, #16 * 13]\n" + "stp x28, x29, [%3, #16 * 14]\n" + "str x30, [%3, #16 * 15]\n" + "mov %0, sp\n" + "adr %1, 1f\n" + "mrs %2, spsr_el1\n" + "1:" + : "=r" (newregs->sp), + "=r" (newregs->pc),Quoted 8 lines
arch/arm64/include/asm/kexec.h line 74
Quoted 4 lines
arch/arm64/include/asm/kexec.h +4
-0+ "=r" (newregs->pstate) + : "r" (&newregs->regs) + : "memory" + );I wonder how useful this thing is, given that it starts by corrupting whatever register is holding
newregs->regs. Maybe this is not supposed to be accurate anyway...Quoted 4 lines
I'm not quite sure about what part of my code you're mentioning here, but
crash_setup_regs()is solely called bycrash_kexec(), andpanic()is the only caller ofcrash_kexec()with NULL argument which, in turn, is used as 'oldregs' incrash_setup_regs().Quoted 6 lines
You have this assembly sequence:
stp x0, x1, [%3, #16 * 0]
[...]where %3 itself is one of the x[0..30] registers.
Not only %3, but also
Quoted 4 lines
So you are saving
things that have already been corrupted by the saving procedure. Not sure how useful that is, but as I said, maybe it is not supposed to be completely accurate.
Quoted 6 lines
x0, x1 ... are the current values in
panic(), and not the exact cpu contexts at the place we are really interested in.
We have no way here inpanic()to know them, but sp and pc would still be useful for back-tracing in later investigation of dump file.Please note that the same problem exists on arm (and x86) implementation.
As I said: if people don't expect to have a precise dump of the register file, then fine.
Quoted 124 lines
Quoted 124 lines
Quoted 124 lines
Given this fact, I think that the values saved in newregs as indicated above will be the best estimate of current cpu contexts.
The other caller of
crash_kexec()isdie()intraps.c, but here we call it with explicit cpu contexts at exception.Quoted 105 lines
Quoted 91 lines
arch/arm64/include/asm/kexec.h +1
-0+ } }#endif /* !defined(__ASSEMBLY__) */ arch/arm64/kernel/machine_kexec.c +29
-2@@ -9,6 +9,7 @@* published by the Free Software Foundation. */ +#include <linux/kernel.h> #include <linux/kexec.h> #include <linux/of_fdt.h> #include <linux/slab.h> @@ -23,6 +24,7 @@extern const unsigned char arm64_relocate_new_kernel[];extern const unsigned long arm64_relocate_new_kernel_size; +bool in_crash_kexec;static unsigned long kimage_start;/** @@ -203,13 +205,38 @@ void machine_kexec(struct kimage *kimage)*/ cpu_soft_restart (virt_to_phys (cpu_reset), -is_hyp_mode_available (), + in_crash_kexec ?0 :is_hyp_mode_available (), reboot_code_buffer_phys, kimage->head, kimage_start);BUG ();/* Should never get here. */ } +/** +* machine_crash_shutdown - shutdown non-boot cpus and save registers +*/ void machine_crash_shutdown (struct pt_regs *regs) { -/* Empty routine needed to avoid build errors. */ +struct pt_regs dummy_regs; +int cpu; + +local_irq_disable (); + + in_crash_kexec = true; + +/* +* clear and initialize the per-cpu info. This is necessary +* because, otherwise, slots for offline cpus would never be +* filled up. See smp_send_stop(). +*/ +memset (&dummy_regs,0 ,sizeof (dummy_regs)); +for_each_possible_cpu (cpu) +crash_save_cpu (&dummy_regs, cpu); + +/* shutdown non-boot cpus */ +smp_send_stop (); + +/* for boot cpu */ +crash_save_cpu (regs,smp_processor_id ()); + +pr_info ("Starting crashdump kernel...\n" ); }arch/arm64/kernel/smp.c +3
-0@@ -37,6 +37,7 @@#include <linux/completion.h> #include <linux/of.h> #include <linux/irq_work.h> +#include <linux/kexec.h> #include <asm/alternative.h> #include <asm/atomic.h> @@ -54,6 +55,8 @@#include <asm/ptrace.h> #include <asm/virt.h> +#include "cpu-reset.h" +#define CREATE_TRACE_POINTS #include <trace/events/ipi.h> @@ -683,8 +686,12 @@ static DEFINE_RAW_SPINLOCK(stop_lock);/* * ipi_cpu_stop - handle IPI from smp_send_stop() */ arch/arm64/kernel/smp.c static DEFINE_RAW_SPINLOCK(stop_lock); · line 697
Quoted 10 lines
arch/arm64/kernel/smp.c +5
-1-static void ipi_cpu_stop (unsigned int cpu) +static void ipi_cpu_stop (unsigned int cpu,struct pt_regs *regs) { +#ifdef CONFIG_KEXEC +/* printing messages may slow down the shutdown. */ + if (!in_crash_kexec) +#endif if (system_state ==SYSTEM_BOOTING || system_state ==SYSTEM_RUNNING ) {raw_spin_lock (&stop_lock);Irrespective of how useful this change is, how about having a predicate instead? Something like:
Quoted 8 lines
static inline bool
is_in_crash_kexec(void)
{#ifdef CONFIG_KEXEC return in_crash_kexec; #else return false;#endif
}OK, I will take your idea.
Quoted 1 line
located in
machine_kexec.c(making thein_crash_kernelstatic), and then
Quoted 7 lines
Quoted 5 lines
Quoted 2 lines
but cannot make
in_crash_kernelstatic because it is also used in bothsmp.candmachine_kexec.c.
smp.conly reads fromin_crash_kernel(at least from what I can see in this patch), so it should be able to use the accessor.Only if we define the accessor as a real function, not an inline function in a header :)
I'll leave the implementation details in your capable hands. :-)
Thanks,
M.
Signature · 6 lines
Jazz is not dead. It just smells funny... _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 12/16] arm64: kdump: implement machine_crash_shutdown()
In reply to Geoff LevandHi Akashi,
Some of the points which came while discussing with Mark Salter are worth including in v13.
Quoted 1 line · Geoff Levand
From: AKASHI Takahiro <[email protected]>
[...]
Quoted 2 lines
+/**
+ *machine_crash_shutdown- shutdown non-boot cpus and save registers
"non-panic" would be correct in stead of "non-boot".
arch/arm64/kernel/machine_kexec.c void machine_kexec(struct kimage *kimage) · line 235
Quoted 1 line
arch/arm64/kernel/machine_kexec.c +1
-0+/* shutdown non-boot cpus */
Ditto
arch/arm64/kernel/machine_kexec.c void machine_kexec(struct kimage *kimage) · line 238
Quoted 3 lines
arch/arm64/kernel/machine_kexec.c +3
-0+smp_send_stop (); + +/* for boot cpu */
"for panic cpu"
[...]
arch/arm64/kernel/smp.c static void ipi_cpu_stop(unsigned int cpu) · line 705
Quoted 3 lines
arch/arm64/kernel/smp.c +0
-0@@ -697,6 +704,11 @@ static void ipi_cpu_stop(unsigned int cpu)local_irq_disable ();
We have "set_cpu_online(cpu, false);" just before it.
Panic core is waiting for non-panic to go offline, i.e. for the above event.
arch/arm64/kernel/smp.c static void ipi_cpu_stop(unsigned int cpu) · line 710
Quoted 5 lines
arch/arm64/kernel/smp.c +4
-0+#ifdef CONFIG_KEXEC +if (in_crash_kexec) +crash_save_cpu (regs, cpu); +#endif /* CONFIG_KEXEC */
However, we are still saving crash info for non-panic core. So, it would be good to move crash_save_cpu() before set_cpu_online() to avoid any race condition.
~Pratyush
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 12/16] arm64: kdump: implement machine_crash_shutdown()
Pratyush,
Quoted 6 lines · Pratyush Anand
Hi Akashi,
Some of the points which came while discussing with Mark Salter are worth including in v13.
On 24/11/2015:10:25:34 PM, Geoff Levand wrote:
Quoted 8 lines
Quoted 1 line
From: AKASHI Takahiro <[email protected]>
[...]
Quoted 2 lines
+/**
+ *machine_crash_shutdown- shutdown non-boot cpus and save registers"non-panic" would be correct in stead of "non-boot".
OK, but for consistency with other places (say, arm/kernel/machine_kexec.c) I prefer "non-crashing cpus."
Quoted 9 lines
arch/arm64/kernel/machine_kexec.c void machine_kexec(struct kimage *kimage) · line 235
Quoted 1 line
arch/arm64/kernel/machine_kexec.c +1
-0+/* shutdown non-boot cpus */ Ditto
arch/arm64/kernel/machine_kexec.c void machine_kexec(struct kimage *kimage) · line 238
Quoted 3 lines
arch/arm64/kernel/machine_kexec.c +3
-0+smp_send_stop (); + +/* for boot cpu */ "for panic cpu"
Ditto.
"for crashing cpu"
Quoted 12 lines
[...]
arch/arm64/kernel/smp.c static void ipi_cpu_stop(unsigned int cpu) · line 705
Quoted 3 lines
arch/arm64/kernel/smp.c +0
-0@@ -697,6 +704,11 @@ static void ipi_cpu_stop(unsigned int cpu)local_irq_disable ();We have "
set_cpu_online(cpu, false);" just before it.
Panic core is waiting for non-panic to go offline, i.e. for the above event.Quoted 1 line
Quoted 7 lines
arch/arm64/kernel/smp.c static void ipi_cpu_stop(unsigned int cpu) · line 710
Quoted 4 lines
arch/arm64/kernel/smp.c +4
-0+#ifdef CONFIG_KEXEC +if (in_crash_kexec) +crash_save_cpu (regs, cpu); +#endif /* CONFIG_KEXEC */ However, we are still saving crash info for non-panic core. So, it would be good to move
crash_save_cpu()beforeset_cpu_online()to avoid any race condition.
Good point, race will be very unlikely though. I will fix it.
Thanks,
-Takahiro AKASHI
Quoted 2 lines
~Pratyush
Re: [PATCH v12 12/16] arm64: kdump: implement machine_crash_shutdown()
Quoted 6 lines · AKASHI Takahiro
Quoted 3 lines
Quoted 1 line
+ *
machine_crash_shutdown- shutdown non-boot cpus and save registers"non-panic" would be correct in stead of "non-boot".
OK, but for consistency with other places (say,
arm/kernel/machine_kexec.c) I prefer "non-crashing cpus."
OK.
Quoted 5 lines
Quoted 5 lines
arch/arm64/kernel/machine_kexec.c void machine_kexec(struct kimage *kimage) · line 235
Quoted 2 lines
arch/arm64/kernel/machine_kexec.c +1
-0+/* shutdown non-boot cpus */ Ditto
Quoted 8 lines
Quoted 5 lines
arch/arm64/kernel/machine_kexec.c void machine_kexec(struct kimage *kimage) · line 238
Quoted 3 lines
arch/arm64/kernel/machine_kexec.c +3
-0+smp_send_stop (); + +/* for boot cpu */ "for panic cpu"
Ditto.
"for crashing cpu"
OK.
~Pratyush
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 13/16 arm64: kdump: add kdump support
In reply to Geoff LevandFrom: AKASHI Takahiro <[email protected]>
On crash dump kernel, all the information about primary kernel's core image is available in elf core header specified by "elfcorehdr=" boot parameter. reserve_elfcorehdr() will set aside the region to avoid any corruption by crash dump kernel.
Crash dump kernel will access the system memory of primary kernel via copy_oldmem_page(), which reads one page by ioremap'ing it since it does not reside in linear mapping on crash dump kernel.
Please note that we should add "mem=X[MG]" boot parameter to limit the memory size and avoid the following assertion at ioremap():
if (WARN_ON(pfn_valid(__phys_to_pfn(phys_addr)))) return NULL;
when accessing any pages beyond the usable memories of crash dump kernel.
We also need our own elfcorehdr_read() here since the weak definition of elfcorehdr_read() utilizes copy_oldmem_page() and will hit the assertion above on arm64.
--- arch/arm64/Kconfig | 12 +++++++ arch/arm64/kernel/Makefile | 1 + arch/arm64/kernel/crash_dump.c | 71 ++++++++++++++++++++++++++++++++++++++++++ arch/arm64/mm/init.c | 29 +++++++++++++++++ 4 files changed, 113 insertions(+) create mode 100644 arch/arm64/kernel/crash_dump.c
arch/arm64/Kconfig +12 -0
@@ -545,6 +545,18 @@ config KEXEC but it is independent of the system firmware. And like a reboot you can start any kernel with it, not just Linux. +config CRASH_DUMP + bool "Build kdump crash kernel" + help + Generate crash dump after being started by kexec. This should + be normally only set in special crash dump kernels which are + loaded in the main kernel with kexec-tools into a specially + reserved region and then later executed after a crash by + kdump/kexec. The crash dump kernel must be compiled to a + memory address not used by the main kernel. + + For more details see Documentation/kdump/kdump.txt + config XEN_DOM0 def_bool y depends on XEN
arch/arm64/kernel/Makefile +1 -0
@@ -43,6 +43,7 @@ arm64-obj-$(CONFIG_ARMV8_DEPRECATED) += armv8_deprecated.o arm64-obj-$(CONFIG_ACPI) += acpi.o arm64-obj-$(CONFIG_KEXEC) += machine_kexec.o relocate_kernel.o \ cpu-reset.o +arm64-obj-$(CONFIG_CRASH_DUMP) += crash_dump.o obj-y += $(arm64-obj-y) vdso/ obj-m += $(arm64-obj-m)
arch/arm64/kernel/crash_dump.c new file +71 -0
71 lines not on this page · raw message
arch/arm64/mm/init.c +29 -0
@@ -35,6 +35,7 @@#include <linux/efi.h> #include <linux/swiotlb.h> #include <linux/kexec.h> +#include <linux/crash_dump.h> #include <asm/fixmap.h> #include <asm/memory.h>
@@ -116,6 +117,31 @@ static void __init reserve_crashkernel(void) }#endif /* CONFIG_KEXEC */ +#ifdef CONFIG_CRASH_DUMP +/* +* reserve_elfcorehdr() - reserves memory for elf core header +* +* This function reserves elf core header given in "elfcorehdr=" kernel +* command line parameter. This region contains all the information about +* primary kernel's core image and is used by a dump capture kernel to +* access the system memory on primary kernel. +*/ +static void __init reserve_elfcorehdr (void ) +{ +if (!elfcorehdr_size) +return ; + +if (memblock_is_region_reserved (elfcorehdr_addr, elfcorehdr_size)) { +pr_warn ("elfcorehdr is overlapped\n" ); +return ; + } + +memblock_reserve (elfcorehdr_addr, elfcorehdr_size); + +pr_info ("Reserving %lldKB of memory at %lldMB for elfcorehdr\n" , + elfcorehdr_size >>10 , elfcorehdr_addr >>20 ); +} +#endif /* CONFIG_CRASH_DUMP */ /* * Return the maximum physical address for ZONE_DMA (DMA_BIT_MASK(32)). It * currently assumes that for memory starting above 4G, 32-bit devices will
@@ -224,6 +250,9 @@ void __init arm64_memblock_init(void)#ifdef CONFIG_KEXEC reserve_crashkernel ();#endif +#ifdef CONFIG_CRASH_DUMP +reserve_elfcorehdr (); +#endif early_init_fdt_scan_reserved_mem ();
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 13/16] arm64: kdump: add kdump support
Quoted 36 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
On crash dump kernel, all the information about primary kernel's core image is available in elf core header specified by "elfcorehdr=" boot parameter.
reserve_elfcorehdr()will set aside the region to avoid any corruption by crash dump kernel.Crash dump kernel will access the system memory of primary kernel via
copy_oldmem_page(), which reads one page by ioremap'ing it since it does not reside in linear mapping on crash dump kernel.
Please note that we should add "mem=X[MG]" boot parameter to limit the memory size and avoid the following assertion atioremap():if (WARN_ON(pfn_valid(__phys_to_pfn(phys_addr)))) return NULL;when accessing any pages beyond the usable memories of crash dump kernel.
We also need our own
elfcorehdr_read()here since the weak definition ofelfcorehdr_read()utilizescopy_oldmem_page()and will hit the assertion above on arm64.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm64/Kconfig | 12 +++++++ arch/arm64/kernel/Makefile | 1 + arch/arm64/kernel/crash_dump.c | 71 ++++++++++++++++++++++++++++++++++++++++++ arch/arm64/mm/init.c | 29 +++++++++++++++++ 4 files changed, 113 insertions(+) create mode 100644 arch/arm64/kernel/crash_dump.carch/arm64/Kconfig +0
-0@@ -545,6 +545,18 @@ config KEXEC but it is independent of the system firmware. And like a reboot you can start any kernel with it, not just Linux.
arch/arm64/Kconfig config KEXEC · line 556
Quoted 10 lines
arch/arm64/Kconfig +9
-0+config CRASH_DUMP + bool "Build kdump crash kernel" + help + Generate crash dump after being started by kexec. This should + be normally only set in special crash dump kernels which are + loaded in the main kernel with kexec-tools into a specially + reserved region and then later executed after a crash by + kdump/kexec. The crash dump kernel must be compiled to a + memory address not used by the main kernel.
What does this even mean? How do I "compile to a memory address not used by the main kernel"?
Quoted 36 lines
arch/arm64/kernel/crash_dump.c new file +30
-0@@ -0,0 +1,71 @@ +/* +* Routines for doing kexec-based kdump +* +* Copyright (C) 2014 Linaro Limited +* Author: AKASHI Takahiro <[email protected]> +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License version 2 as +* published by the Free Software Foundation. +*/ + +#include <linux/crash_dump.h> +#include <linux/errno.h> +#include <linux/io.h> +#include <linux/memblock.h> +#include <linux/uaccess.h> +#include <asm/memory.h> + +/** +* copy_oldmem_page() - copy one page from old kernel memory +* @pfn: page frame number to be copied +* @buf: buffer where the copied page is placed +* @csize: number of bytes to copy +* @offset: offset in bytes into the page +* @userbuf: if set, @buf is in a user address space +* +* This function copies one page from old kernel memory into buffer pointed by +* @buf. If @buf is in userspace, set @userbuf to %1. Returns number of bytes +* copied or negative error in case of failure. +*/
arch/arm64/kernel/crash_dump.c line 40
Quoted 10 lines
arch/arm64/kernel/crash_dump.c new file +10
-0+ssize_t copy_oldmem_page (unsigned long pfn,char *buf, +size_t csize,unsigned long offset, +int userbuf) +{ +void *vaddr; + +if (!csize) +return 0 ; + + vaddr =ioremap_cache (pfn <<PAGE_SHIFT ,PAGE_SIZE );
pfn_to_page
Quoted 21 lines
arch/arm64/kernel/crash_dump.c new file +21
-0+if (!vaddr) +return -ENOMEM ; + +if (userbuf) { +if (copy_to_user (buf, vaddr + offset, csize)) { +iounmap (vaddr); +return -EFAULT ; + } + }else { +memcpy (buf, vaddr + offset, csize); + } + +iounmap (vaddr); + +return csize; +} + +/** +* elfcorehdr_read - read from ELF core header +* @buf: buffer where the data is placed +* @csize: number of bytes to read
arch/arm64/kernel/crash_dump.c line 71
Quoted 10 lines
arch/arm64/kernel/crash_dump.c new file +10
-0+* @ppos: address in the memory +* +* This function reads @count bytes from elf core header which exists +* on crash dump kernel's memory. +*/ +ssize_t elfcorehdr_read (char *buf,size_t count,u64 *ppos) +{ +memcpy (buf,phys_to_virt ((phys_addr_t)*ppos), count); +return count; +}
I know you say that we have to override this function so that we don't hit the pfn_valid warning in ioremap, but what guarantees that the ELF header of the crashed kernel is actually mapped in our linear mapping?
Quoted 29 lines
arch/arm64/mm/init.c +14
-0@@ -35,6 +35,7 @@#include <linux/efi.h> #include <linux/swiotlb.h> #include <linux/kexec.h> +#include <linux/crash_dump.h> #include <asm/fixmap.h> #include <asm/memory.h> @@ -116,6 +117,31 @@ static void __init reserve_crashkernel(void) }#endif /* CONFIG_KEXEC */ +#ifdef CONFIG_CRASH_DUMP +/* +* reserve_elfcorehdr() - reserves memory for elf core header +* +* This function reserves elf core header given in "elfcorehdr=" kernel +* command line parameter. This region contains all the information about +* primary kernel's core image and is used by a dump capture kernel to +* access the system memory on primary kernel. +*/ +static void __init reserve_elfcorehdr (void ) +{ +if (!elfcorehdr_size) +return ;
arch/arm64/mm/init.c static void __init reserve_crashkernel(void) · line 142
Quoted 10 lines
arch/arm64/mm/init.c +10
-0+ +if (memblock_is_region_reserved (elfcorehdr_addr, elfcorehdr_size)) { +pr_warn ("elfcorehdr is overlapped\n" ); +return ; + } + +memblock_reserve (elfcorehdr_addr, elfcorehdr_size); + +pr_info ("Reserving %lldKB of memory at %lldMB for elfcorehdr\n" , + elfcorehdr_size >>10 , elfcorehdr_addr >>20 );
I'd have thought it would be more useful to print the address as an address rather than a size.
arch/arm64/mm/init.c static void __init reserve_crashkernel(void) · line 143
Quoted 1 line
arch/arm64/mm/init.c +1
-0+}
Similar #else trick here.
Will
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 13/16] arm64: kdump: add kdump support
Quoted 38 lines · Will Deacon
Quoted 37 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
On crash dump kernel, all the information about primary kernel's core image is available in elf core header specified by "elfcorehdr=" boot parameter.
reserve_elfcorehdr()will set aside the region to avoid any corruption by crash dump kernel.Crash dump kernel will access the system memory of primary kernel via
copy_oldmem_page(), which reads one page by ioremap'ing it since it does not reside in linear mapping on crash dump kernel.
Please note that we should add "mem=X[MG]" boot parameter to limit the memory size and avoid the following assertion atioremap():if (WARN_ON(pfn_valid(__phys_to_pfn(phys_addr)))) return NULL;when accessing any pages beyond the usable memories of crash dump kernel.
We also need our own
elfcorehdr_read()here since the weak definition ofelfcorehdr_read()utilizescopy_oldmem_page()and will hit the assertion above on arm64.Signed-off-by AKASHI Takahiro <[email protected]>--- arch/arm64/Kconfig | 12 +++++++ arch/arm64/kernel/Makefile | 1 + arch/arm64/kernel/crash_dump.c | 71 ++++++++++++++++++++++++++++++++++++++++++ arch/arm64/mm/init.c | 29 +++++++++++++++++ 4 files changed, 113 insertions(+)create mode 100644
arch/arm64/kernel/crash_dump.carch/arm64/Kconfig +0
-0@@ -545,6 +545,18 @@ config KEXEC but it is independent of the system firmware. And like a reboot you can start any kernel with it, not just Linux.
Quoted 12 lines
arch/arm64/Kconfig config KEXEC · line 556
Quoted 9 lines
arch/arm64/Kconfig +9
-0+config CRASH_DUMP + bool "Build kdump crash kernel" + help + Generate crash dump after being started by kexec. This should + be normally only set in special crash dump kernels which are + loaded in the main kernel with kexec-tools into a specially + reserved region and then later executed after a crash by + kdump/kexec. The crash dump kernel must be compiled to a + memory address not used by the main kernel.What does this even mean? How do I "compile to a memory address not used by the main kernel"?
Well, it's just a copy from arm, but right, it's ambiguous. I will remove that text.
Quoted 40 lines
Quoted 40 lines
arch/arm64/kernel/crash_dump.c new file +34
-0@@ -0,0 +1,71 @@ +/* +* Routines for doing kexec-based kdump +* +* Copyright (C) 2014 Linaro Limited +* Author: AKASHI Takahiro <[email protected]> +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License version 2 as +* published by the Free Software Foundation. +*/ + +#include <linux/crash_dump.h> +#include <linux/errno.h> +#include <linux/io.h> +#include <linux/memblock.h> +#include <linux/uaccess.h> +#include <asm/memory.h> + +/** +* copy_oldmem_page() - copy one page from old kernel memory +* @pfn: page frame number to be copied +* @buf: buffer where the copied page is placed +* @csize: number of bytes to copy +* @offset: offset in bytes into the page +* @userbuf: if set, @buf is in a user address space +* +* This function copies one page from old kernel memory into buffer pointed by +* @buf. If @buf is in userspace, set @userbuf to %1. Returns number of bytes +* copied or negative error in case of failure. +*/ +ssize_t copy_oldmem_page (unsigned long pfn,char *buf, +size_t csize,unsigned long offset, +int userbuf) +{
Quoted 8 lines
arch/arm64/kernel/crash_dump.c line 40
Quoted 6 lines
arch/arm64/kernel/crash_dump.c new file +6
-0+void *vaddr; + +if (!csize) +return 0 ; + + vaddr =ioremap_cache (pfn <<PAGE_SHIFT ,PAGE_SIZE );
pfn_to_page
Maybe __pfn_to_phsy()?
Quoted 27 lines
Quoted 27 lines
arch/arm64/kernel/crash_dump.c new file +27
-0+if (!vaddr) +return -ENOMEM ; + +if (userbuf) { +if (copy_to_user (buf, vaddr + offset, csize)) { +iounmap (vaddr); +return -EFAULT ; + } + }else { +memcpy (buf, vaddr + offset, csize); + } + +iounmap (vaddr); + +return csize; +} + +/** +* elfcorehdr_read - read from ELF core header +* @buf: buffer where the data is placed +* @csize: number of bytes to read +* @ppos: address in the memory +* +* This function reads @count bytes from elf core header which exists +* on crash dump kernel's memory. +*/ +ssize_t elfcorehdr_read (char *buf,size_t count,u64 *ppos)
Quoted 8 lines
arch/arm64/kernel/crash_dump.c line 71
Quoted 4 lines
arch/arm64/kernel/crash_dump.c new file +4
-0+{ +memcpy (buf,phys_to_virt ((phys_addr_t)*ppos), count); +return count; +}I know you say that we have to override this function so that we don't hit the
pfn_validwarning in ioremap, but what guarantees that the ELF header of the crashed kernel is actually mapped in our linear mapping?
Well, in fact, it depends on kexec-tools.
In the current implementation for arm64, the elf core header is allocated within the usable memory of crash dump kernel.
Should we add some check here?
Quoted 34 lines
Quoted 34 lines
arch/arm64/mm/init.c +19
-0@@ -35,6 +35,7 @@#include <linux/efi.h> #include <linux/swiotlb.h> #include <linux/kexec.h> +#include <linux/crash_dump.h> #include <asm/fixmap.h> #include <asm/memory.h> @@ -116,6 +117,31 @@ static void __init reserve_crashkernel(void) }#endif /* CONFIG_KEXEC */ +#ifdef CONFIG_CRASH_DUMP +/* +* reserve_elfcorehdr() - reserves memory for elf core header +* +* This function reserves elf core header given in "elfcorehdr=" kernel +* command line parameter. This region contains all the information about +* primary kernel's core image and is used by a dump capture kernel to +* access the system memory on primary kernel. +*/ +static void __init reserve_elfcorehdr (void ) +{ +if (!elfcorehdr_size) +return ; + +if (memblock_is_region_reserved (elfcorehdr_addr, elfcorehdr_size)) { +pr_warn ("elfcorehdr is overlapped\n" ); +return ; + }
Quoted 8 lines
arch/arm64/mm/init.c static void __init reserve_crashkernel(void) · line 142
Quoted 5 lines
arch/arm64/mm/init.c +5
-0+ +memblock_reserve (elfcorehdr_addr, elfcorehdr_size); + +pr_info ("Reserving %lldKB of memory at %lldMB for elfcorehdr\n" , + elfcorehdr_size >>10 , elfcorehdr_addr >>20 );I'd have thought it would be more useful to print the address as an address rather than a size.
Yeah, I totally agree, but all the other archs, including x86 and arm, print the address in "%lldMB" format.
If you like, I can fix it.
Quoted 3 lines
arch/arm64/mm/init.c static void __init reserve_crashkernel(void) · line 143
Quoted 1 line
arch/arm64/mm/init.c +1
-0+}Similar #else trick here.
Sure.
Thanks,
-Takahiro AKASHI
Quoted 2 lines
Will
PATCH v12 14/16 arm64: kdump: update a kernel doc
In reply to Geoff LevandFrom: AKASHI Takahiro <[email protected]>
This patch adds arch specific descriptions about kdump usage on arm64 to kdump.txt.
--- Documentation/kdump/kdump.txt | 32 +++++++++++++++++++++++++++++++- 1 file changed, 31 insertions(+), 1 deletion(-)
Documentation/kdump/kdump.txt +31 -1
@@ -18,7 +18,7 @@ memory image to a dump file on the local disk, or across the network to a remote system. Kdump and kexec are currently supported on the x86, x86_64, ppc64, ia64, -s390x and arm architectures. +s390x, arm and arm64 architectures. When the system kernel boots, it reserves a small section of memory for the dump-capture kernel. This ensures that ongoing Direct Memory Access
@@ -249,6 +249,29 @@ Dump-capture kernel config options (Arch Dependent, arm) AUTO_ZRELADDR=y +Dump-capture kernel config options (Arch Dependent, arm64) +---------------------------------------------------------- + +1) Disable symmetric multi-processing support under "Processor type and + features": + + CONFIG_SMP=n + + (If CONFIG_SMP=y, then specify maxcpus=1 on the kernel command line + when loading the dump-capture kernel, see section "Load the Dump-capture + Kernel".) + +2) The maximum memory size on the dump-capture kernel must be limited by + specifying: + + mem=X[MG] + + where X should be less than or equal to the size in "crashkernel=" + boot parameter. Kexec-tools will automatically add this. + +3) Currently, kvm will not be enabled on the dump-capture kernel even + if it is configured. + Extended crashkernel syntax ===========================
@@ -312,6 +335,8 @@ Boot into System Kernel any space below the alignment point may be overwritten by the dump-capture kernel, which means it is possible that the vmcore is not that precise as expected. + On arm64, use "crashkernel=Y[@X]". Note that the start address of + the kernel, X if explicitly specified, must be aligned to 2MiB (0x200000). Load the Dump-capture Kernel ============================
@@ -334,6 +359,8 @@ For s390x: - Use image or bzImage For arm: - Use zImage +For arm64: + - Use vmlinux If you are using a uncompressed vmlinux image then use following command to load dump-capture kernel.
@@ -377,6 +404,9 @@ For s390x: For arm: "1 maxcpus=1 reset_devices" +For arm64: + "1 mem=X[MG] maxcpus=1 reset_devices" + Notes on loading the dump-capture kernel: * By default, the ELF headers are stored in ELF64 format to support
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 14/16] arm64: kdump: update a kernel doc
Quoted 28 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
This patch adds arch specific descriptions about kdump usage on arm64 to kdump.txt.
Signed-off-by AKASHI Takahiro <[email protected]>--- Documentation/kdump/kdump.txt | 32 +++++++++++++++++++++++++++++++- 1 file changed, 31 insertions(+), 1 deletion(-)Documentation/kdump/kdump.txt +2
-1@@ -18,7 +18,7 @@ memory image to a dump file on the local disk, or across the network to a remote system. Kdump and kexec are currently supported on the x86, x86_64, ppc64, ia64, -s390x and arm architectures. +s390x, arm and arm64 architectures. When the system kernel boots, it reserves a small section of memory for the dump-capture kernel. This ensures that ongoing Direct Memory Access@@ -249,6 +249,29 @@ Dump-capture kernel config options (Arch Dependent, arm) AUTO_ZRELADDR=y +Dump-capture kernel config options (Arch Dependent, arm64)
Documentation/kdump/kdump.txt Dump-capture kernel config options (Arch Dependent, arm) · line 262
Quoted 10 lines
Documentation/kdump/kdump.txt +10
-0+---------------------------------------------------------- + +1) Disable symmetric multi-processing support under "Processor type and + features": + + CONFIG_SMP=n + + (If CONFIG_SMP=y, then specify maxcpus=1 on the kernel command line + when loading the dump-capture kernel, see section "Load the Dump-capture + Kernel".)
SMP is now forced on, so please update this text.
Quoted 24 lines
Documentation/kdump/kdump.txt +13
-0+2) The maximum memory size on the dump-capture kernel must be limited by + specifying: + + mem=X[MG] + + where X should be less than or equal to the size in "crashkernel=" + boot parameter. Kexec-tools will automatically add this. + +3) Currently, kvm will not be enabled on the dump-capture kernel even + if it is configured. + Extended crashkernel syntax ===========================@@ -312,6 +335,8 @@ Boot into System Kernel any space below the alignment point may be overwritten by the dump-capture kernel, which means it is possible that the vmcore is not that precise as expected. + On arm64, use "crashkernel=Y[@X]". Note that the start address of + the kernel, X if explicitly specified, must be aligned to 2MiB (0x200000). Load the Dump-capture Kernel ============================
Documentation/kdump/kdump.txt For s390x: · line 363
Quoted 5 lines
Documentation/kdump/kdump.txt +2
-0@@ -334,6 +359,8 @@ For s390x: - Use image or bzImage For arm: - Use zImage +For arm64: + - Use vmlinux
What? Why doesn't Image work?
Will
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 14/16] arm64: kdump: update a kernel doc
Quoted 28 lines · Will Deacon
Quoted 27 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
This patch adds arch specific descriptions about kdump usage on arm64 to kdump.txt.
Signed-off-by AKASHI Takahiro <[email protected]>--- Documentation/kdump/kdump.txt | 32 +++++++++++++++++++++++++++++++- 1 file changed, 31 insertions(+), 1 deletion(-)Documentation/kdump/kdump.txt +1
-1@@ -18,7 +18,7 @@ memory image to a dump file on the local disk, or across the network to a remote system. Kdump and kexec are currently supported on the x86, x86_64, ppc64, ia64, -s390x and arm architectures. +s390x, arm and arm64 architectures. When the system kernel boots, it reserves a small section of memory for the dump-capture kernel. This ensures that ongoing Direct Memory Access@@ -249,6 +249,29 @@ Dump-capture kernel config options (Arch Dependent, arm) AUTO_ZRELADDR=y
Quoted 13 lines
Documentation/kdump/kdump.txt Dump-capture kernel config options (Arch Dependent, arm) · line 262
Quoted 11 lines
Documentation/kdump/kdump.txt +11
-0+Dump-capture kernel config options (Arch Dependent, arm64) +---------------------------------------------------------- + +1) Disable symmetric multi-processing support under "Processor type and + features": + + CONFIG_SMP=n + + (If CONFIG_SMP=y, then specify maxcpus=1 on the kernel command line + when loading the dump-capture kernel, see section "Load the Dump-capture + Kernel".)SMP is now forced on, so please update this text.
OK, I will fix it.
Quoted 21 lines
Quoted 21 lines
Documentation/kdump/kdump.txt +13
-0+2) The maximum memory size on the dump-capture kernel must be limited by + specifying: + + mem=X[MG] + + where X should be less than or equal to the size in "crashkernel=" + boot parameter. Kexec-tools will automatically add this. + +3) Currently, kvm will not be enabled on the dump-capture kernel even + if it is configured. + Extended crashkernel syntax ===========================@@ -312,6 +335,8 @@ Boot into System Kernel any space below the alignment point may be overwritten by the dump-capture kernel, which means it is possible that the vmcore is not that precise as expected. + On arm64, use "crashkernel=Y[@X]". Note that the start address of + the kernel, X if explicitly specified, must be aligned to 2MiB (0x200000).
Quoted 10 lines
Documentation/kdump/kdump.txt For s390x: · line 363
Quoted 8 lines
Documentation/kdump/kdump.txt +2
-0Load the Dump-capture Kernel ============================@@ -334,6 +359,8 @@ For s390x: - Use image or bzImage For arm: - Use zImage +For arm64: + - Use vmlinuxWhat? Why doesn't Image work?
Actually it depends on kexec-tools, not the kernel. Initially Geoff implemented vmlinux support only, then recently Pratyush (RedHat) gave us a patch for "Image" support.
So now Image should work.
-Takahiro AKASHI
Quoted 2 lines
Will
PATCH v12 15/16 arm64: kdump: enable kdump in the arm64 defconfig
In reply to Geoff LevandFrom: AKASHI Takahiro <[email protected]>
--- arch/arm64/configs/defconfig | 1 + 1 file changed, 1 insertion(+)
arch/arm64/configs/defconfig +1 -0
@@ -59,6 +59,7 @@ CONFIG_KSM=y CONFIG_TRANSPARENT_HUGEPAGE=y CONFIG_CMA=y CONFIG_KEXEC=y +CONFIG_CRASH_DUMP=y CONFIG_CMDLINE="console=ttyAMA0" # CONFIG_CORE_DUMP_DEFAULT_ELF_HEADERS is not set CONFIG_COMPAT=y
Signature · 8 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
PATCH v12 16/16 arm64: kdump: relax BUG_ON() if more than one cpus are still active
In reply to Geoff LevandFrom: AKASHI Takahiro <[email protected]>
We should try best in case of kdump.
So even if not all secondary cpus have shut down, we do kdump anyway.
--- arch/arm64/kernel/machine_kexec.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)
arch/arm64/kernel/machine_kexec.c +7 -1
@@ -148,7 +148,13 @@ void machine_kexec(struct kimage *kimage)phys_addr_t reboot_code_buffer_phys;void *reboot_code_buffer; -BUG_ON (num_online_cpus () >1 ); +if (num_online_cpus () >1 ) { +if (in_crash_kexec) +pr_warn ("kdump might fail because %d cpus are still online\n" , +num_online_cpus ()); +else +BUG (); + } reboot_code_buffer_phys =page_to_phys (kimage->control_code_page); reboot_code_buffer =phys_to_virt (reboot_code_buffer_phys);
Signature · 7 lines
2.5.0 _______________________________________________ kexec mailing list [email protected] http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 16/16] arm64: kdump: relax BUG_ON() if more than one cpus are still active
Quoted 14 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
We should try best in case of kdump.
So even if not all secondary cpus have shut down, we do kdump anyway.--- arch/arm64/kernel/machine_kexec.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)arch/arm64/kernel/machine_kexec.c +0
-0@@ -148,7 +148,13 @@ void machine_kexec(struct kimage *kimage)phys_addr_t reboot_code_buffer_phys;
arch/arm64/kernel/machine_kexec.c void machine_kexec(struct kimage *kimage) · line 157
Quoted 10 lines
arch/arm64/kernel/machine_kexec.c +7
-1void *reboot_code_buffer; -BUG_ON (num_online_cpus () >1 ); +if (num_online_cpus () >1 ) { +if (in_crash_kexec) +pr_warn ("kdump might fail because %d cpus are still online\n" , +num_online_cpus ()); +else +BUG (); + }
Can you just rewrite the existing BUG_ON as:
BUG_ON(num_online_cpus() && !WARN_ON(in_crash_kexec));
?
Will
_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
Re: [PATCH v12 16/16] arm64: kdump: relax BUG_ON() if more than one cpus are still active
Quoted 28 lines · Will Deacon
Quoted 14 lines · Geoff Levand
From: AKASHI Takahiro <[email protected]>
We should try best in case of kdump.
So even if not all secondary cpus have shut down, we do kdump anyway.--- arch/arm64/kernel/machine_kexec.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)arch/arm64/kernel/machine_kexec.c +0
-0@@ -148,7 +148,13 @@ void machine_kexec(struct kimage *kimage)phys_addr_t reboot_code_buffer_phys;arch/arm64/kernel/machine_kexec.c void machine_kexec(struct kimage *kimage) · line 157
Quoted 10 lines
arch/arm64/kernel/machine_kexec.c +7
-1void *reboot_code_buffer; -BUG_ON (num_online_cpus () >1 ); +if (num_online_cpus () >1 ) { +if (in_crash_kexec) +pr_warn ("kdump might fail because %d cpus are still online\n" , +num_online_cpus ()); +else +BUG (); + }Can you just rewrite the existing
BUG_ONas:
Quoted 3 lines
BUG_ON(num_online_cpus() && !WARN_ON(in_crash_kexec));?
Yes, but
BUG_ON((num_online_cpus() > 1) && !WARN_ON(in_crash_kexec));
Thanks,
-Takahiro AKASHI
Quoted 2 lines
Will